Wikipedia talk:Speedy deletion

See also Wikipedia talk:Speedy deletions for discussions which took place in 2006-07 before it was redirected here.

RfC: Including Markdown in G15

The following discussion is closed. Please do not modify it. Subsequent comments should be made on the appropriate discussion page. No further edits should be made to this discussion.

---

Should the following criterion be added to G15? Chaotic Enby (talk · contribs) 21:44, 4 August 2025 (UTC)

---

• Markdown-formatted text: This may include using Markdown formatting for links ([example](https://example.com)), italics (*example*), or bold (**example**), which is often generated by large language models in place of wikitext.

Background

This is a follow-up on the previous RfC, where the inclusion of Markdown to the criterion was suggested.

Survey (Markdown)

• Support as proposer. Chaotic Enby (talk · contribs) 21:44, 4 August 2025 (UTC)
• Support per my previous argument. Tenshi! (Talk page) 21:45, 4 August 2025 (UTC)

  On needing to be experienced with Wikipedia to know that Markdown doesn't work, we have a preview button which editors can use, which would clearly show that the Markdown format isn't working. I don't necessarily agree that this criterion should be used for 1 instance of someone using - instead of *, but when the entire page is plastered with Markdown syntax, that is clearly unreviewed LLM output. Tenshi! (Talk page) 23:34, 4 August 2025 (UTC)

• Support with qualifier because any human who's actually paying attention would realize that the Markdown formatting does not render as they expect. Also, it has an edit filter. Much like how one nonexistent reference is not definitive in the current criteria, this criterion should not be the only reason for nominating the page. SuperPianoMan9167 (talk) 21:50, 4 August 2025 (UTC)

  I'm actually leaning a bit more towards Neutral or Weak oppose after reading the arguments of Sohom Datta, Thryduulf, and others; yes, Markdown is usually a sign of copy-pasted LLM output, but it's entirely possible that a newbie might paste in a (human-written) draft from a text editor that uses Markdown and not realize that they need to convert it to wikitext, or else they may type in Markdown out of habit. The editors in opposition in this RFC make a very good point that speedy deletion is only for the most obvious cases; perhaps we could treat Markdown as a preemptive sign of LLM writing without explicitly mentioning it in the criterion. When making sure that a page is unquestionably LLM-generated without human review, editors would just consult the other criteria, much like how the edit filter for Markdown must be manually reviewed to catch false positives. The edit filter itself even says [Markdown] is often, but not always indicative of AI-generated text. It's probably a good idea to avoid putting something that isn't a clear, unambiguous sign of LLM usage in the speedy deletion criterion to avoid deleting pages that should not be deleted, but then again, we already have the disclaimer Since humans can make typos and links may suffer from link rot, a single example should not be considered definitive. Editors should use additional methods to verify whether a reference truly does not exist. for fake references in the current CSD. SuperPianoMan9167 (talk) 07:46, 6 August 2025 (UTC)

• Support since most people who actually know how to use Markdown would realize it doesn't work here. Children Will Listen (🐄 talk, 🫘 contribs) 21:56, 4 August 2025 (UTC)

  I use Markdown every single day too, but I definitely won't use it to write an entire article filled with promotional language. Even if I were to accidentally insert, say, a section formatted completely using markdown, I would quickly self-revert or fix it. Children Will Listen (🐄 talk, 🫘 contribs) 22:22, 4 August 2025 (UTC)

• Strong oppose, As a person who has spent a fair bit of time in the developer ecosystem outside Wikipedia, markdown is a very common language to take notes in and write documents. I personally often Obsidian to take notes and draft replies/article passages. I'm 100% sure that there will be cases where I (or insert developer-oriented newbie) will accidentally use markdown syntax or forget to convert their notes from wikitext. WP:CSD is for deletion of unambigous instances of abuse and while I do see unedited markdown as a strong signal of LLM generated text, I would hesitate to call it a unambigous indicator. We should not be catching newbies in this cross-fire. -- Sohom (talk) 22:10, 4 August 2025 (UTC)
• Oppose. In and of itself, Markdown is not strong evidence of AI usage, or at least not nearly as strong as the other bullet points. I occasionally catch myself using Markdown syntax in Wikitext; I usually fix it before I hit publish, but it is absolutely plausible that I would accidentally create a page with some Markdown in it (and I have never used LLMs on Wikipedia). Giraffer (talk) 22:13, 4 August 2025 (UTC)

  I agree with Sohom, especially his bit about the CSD threshold. Giraffer (talk) 22:14, 4 August 2025 (UTC)

• Support basically per ChildrenWillListen; I think Sohom Datta et. al miss the point; it's not just using markdown that would be required here, but using markdown, and not even noticing that it doesn't work. * Pppery * _{it has begun...} 22:16, 4 August 2025 (UTC)

  I do get that, but accidents do happen and as a newbie, you could go "oh maybe I don't have the permission to do this maybe" or "oh I will fix this in my next edit". I think the introduction of something like this as a WP:CSD criteria will develop a culture of folks going "oh this is markdown, click the CSD button" without giving it to much thought. Sohom (talk) 22:32, 4 August 2025 (UTC)

  If the newbie used the preview button, they would have probably realized that Markdown doesn't work on Wikipedia. Besides, most obviously LLM-generated pages with Markdown also have other problems that meet the other criteria for G15, like the "oaicite" code. Maybe this criterion should have a disclaimer saying "Humans may accidentally use Markdown instead of wikitext, so this criterion should not be considered definitive." SuperPianoMan9167 (talk) 22:41, 4 August 2025 (UTC)

  For me and you the preview button is obvious, for newbies it might/will not be. If this proposal was specifically (and narrowly) about oacite and other LLM nonsense like "+19 sites" I would have probably supported it. Also the unfortunate fact of the matter that new patrollers will see the bolded part and the takeaway will be "click the CSD button if you see markdown". (and I don't disagree that they will be right 9 times out of 10, but that 1 is the problem) As I said above, yes, we should consider it a strong indicator, but I'm hesitant to put it as a explicit CSD criteria (a disclaimer is a move in the right direction tho). Sohom (talk) 23:41, 4 August 2025 (UTC)

• Perhaps with a qualifier in the same spirit as "Implausible non-existent references"'s "a single example should not be considered definitive": markdown on its own should not be speedyable, but it can be taken as proof if the page is worded like LLM output. —Cryptic 22:29, 4 August 2025 (UTC)

  +1. Basically all of the obviously LLM-generated pages I've seen have other problems besides Markdown. SuperPianoMan9167 (talk) 22:42, 4 August 2025 (UTC)

  I added this as a suggestion in the "Discussion" section below. SuperPianoMan9167 (talk) 23:17, 4 August 2025 (UTC)

• Weak support - While I understand Sohom's concerns with non-AI users being caught up in this, from what I've seen this isn't very common, or at least not nearly as common as it is when AI adds it. For me markdown is one of the most obvious indicators and the first thing I notice when I see an AI article, so I think this addition will be a net positive for G15. Sophisticatedevening^(talk) 22:49, 4 August 2025 (UTC)

  Yep, Markdown is easy to spot compared to some of the other signs of AI writing; it also has an edit filter for that reason. SuperPianoMan9167 (talk) 23:22, 4 August 2025 (UTC)

• Weakstrong oppose per Sohom. Way before the dawn of ChatGPT have newbies cluelessly tried to use * for italics and bolding without previewing, anecdotally. This isn't something that will work often with the way most AI responses are copied either. As for "Weakstrong", I wanted to qualify my !vote as between a weak oppose and a standard oppose as maybe it should be documented somewhere that currently markdown syntax in new pages is correlated with LLM usage. Aaron Liu (talk) 22:54, 4 August 2025 (UTC)
• Very strong oppose per Sohom. A human will see that markdown hasn't worked, but you need to be experienced with Wikipedia to know why it hasn't worked, and if you don't know why it hasn't worked you don't know that you need to remove it. That combined with the number of times I've seen and experienced markdown not converted to formatting on sites that do use markdown makes it very clear to me that this is absolutely not an unambiguous indication that (a) this is LLM-generated, or (b) that the text has not been reviewed by a human. Thryduulf (talk) 23:01, 4 August 2025 (UTC)

  Maybe Markdown would work as long as it is not the only reason for nominating the page, like how one nonexistent reference is not definitive under the current G15 criteria. Yes, Markdown is not an unambiguous sign of LLM use, but it appears quite commonly alongside other signs of LLM use, and so it should probably be noted as such. SuperPianoMan9167 (talk) 23:05, 4 August 2025 (UTC)

  We're going to have a hard enough time keeping the misuse of this criteria to a minimum as it is, the last thing we want to be doing is to add things that are not a reliable indicator of either LLM-use or of the lack of human review to it. Thryduulf (talk) 00:26, 5 August 2025 (UTC)

• Oppose at least for now. I'd like to see how the more clear-cut criteria play out before considering the necessity of expanding them. -- LWG ^talk 00:34, 5 August 2025 (UTC)
• Oppose as standalone criterion, per Sohom and Thryduulf. I understand that Markdown can be a strong indicator of LLM. However, it can also be an indicator of a human editor who's more familiar with sites like Reddit and Discord, and doesn't know that Markdown formatting will never work on WP. The thought process of people intentionally using Markdown on WP isn't going to be "Hey, Markdown doesn't work, let me use a completely different formatting language like wikitext", it's "Hey, Markdown doesn't work, let me try to change the Markdown formatting until I come across something that does work". Unfortunately, this criterion would catch these types of people. In any case, articles that qualify for G15 usually have a bunch of other telltale signs besides just Markdown, which can just be a sign of a wikicode-illiterate editor. Epicgenius (talk) 03:31, 5 August 2025 (UTC)
• Support provided that the creator of the page had either successfully inserted rich text on Wikipedia at least once (without the use of a preload) or otherwise demonstrated that they knew how to format wikicode correctly (such as cross-wiki experience), had one of their pages deleted for G15, or inserted content that would be G15-eligible if it were a separate page (i.e. prior LLM use: if they abused LLMs in the past, the markdown-formatted page is more likely to also be LLM-generated without review.) prior to the creation of the page. Otherwise, weak oppose per the other oppose voters -- the editor may have thought that they did not have permission to insert rich text due to anti-spam, for one example. OutsideNormality (talk) 04:10, 5 August 2025 (UTC)
• Support with the proviso that this should be an indicator of AI writing and that administrators should continue to use their judgement. Stifle (talk) 10:23, 5 August 2025 (UTC)
• Moral support. The proposed wording is unworkable, as the significant opposition here shows. Markdown can be an AI tell, but it can also be used by good-faith editors, and many have correctly pointed out that not everyone knows how to use the Preview tool (I certainly didn't until I broke something [1] twice [2] as a new user). What might work instead is wording like "Extensive broken markdown formatting", though I still don't know if that alone would be enough for speedy deletion if the page doesn't meet any of the other criteria. Toadspike [Talk] 11:19, 5 August 2025 (UTC)
• Oppose, while LLM's may output Markdown, other tools have added Markdown support recently, for example the Notes app in iOS 26 and iPadOS 26 support exporting as Markdown now. Presumably someone who uses those tools to draft an article or revision may not be fully aware of the formatting differences. I wanna say Notepad in Windows 11 also supports Markdown. —Locke Cole • t • c 22:52, 5 August 2025 (UTC)
• Oppose. While this may be a telltale clue, it's not by itself indicitive. Heck, I've been an admin since Feburary 2011 (to be fair, with some lengthy Wikibreaks) and I'm pretty sure I've accidentally posted comments sometimes where I used markdown instead of wikitext out of muscle-memory habit. - The Bushranger _{One ping only} 03:38, 6 August 2025 (UTC)
• Oppose If an article manages to be largely error free, both in prose and in citations, and the only fault is the use of markdown, it is plausible that the LLM output did undergo human review. CSDs should be only be used in the most obvious cases. I am also convinced by the testimonials by other editors in this discussion of accidental uses of markdown, showing that this is not a foolproof indicator of LLM use. The current guideline does allow the use of markdown as justification; it is just that it should not be the sole criterion. Any borderline cases can be discussed at WP:AfD. Ca ^{talk to me!} 06:49, 6 August 2025 (UTC)
• Oppose I think that the use of markdown, while it is often a sign of AI writing, does not definitively mean that the article in question is the unchecked output of an LLM, so it should not be a criterion for speedy deletion on its own. I am basing this on a half-remembered article that did use a bunch of markdown, was probably generated with AI, but the references and text-source integrity were OK, so I just cleaned it up a bit - it wasn't the best article in the world, but it wasn't speedy deletable. I wouldn't be opposed to reconsidering this position in the future if there was clear evidence that considerable time was still being wasted on irredeemable articles containing markdown at, e.g., AfC or AfD that were ultimately rejected or deleted. Cheers, SunloungerFrog (talk) 09:07, 6 August 2025 (UTC)
• Oppose - loads of people habitually use markdown in various jobs and fields and it is weird (non-intuiative) that it doesn't work here. Not only can someone accidently add markdown without thinking, it's also entirely possible that markdown could or would be supported by future versions of the wikimedia software. JMWt (talk) 09:37, 6 August 2025 (UTC)
• Oppose per Sohom. I tend to be hawkish about AI, but this criterion will very likely lead to misplaced CSDs and some very confused new editors. Cremastra (talk · contribs) 19:49, 6 August 2025 (UTC)
• Oppose per Sohom. I am also sceptical about generative AI, but I feel that markdown syntax may not always indicate or be a sole indicator of an AI-generated article. It takes a bit of time to get used to MediaWiki formatting, which is quite different from HTML and markdown. ----Minoa (talk) 18:09, 8 August 2025 (UTC)
• Strong Oppose: Just Fix It. This sort of markdown to wikitext conversion is literally find and replace a lot of the time (albeit with regular expressions). And even if it's not, having ugly formatting is not a reason to speedy delete an article. lethargilistic (talk) 02:21, 10 August 2025 (UTC)

  @Lethargilistic You're missing the point. The point is that it's an indicator of AI-generated text, which will have more problems which aren't immediately noticeable. It's not an aesthetic issue at all. Cremastra (talk · contribs) 14:26, 11 August 2025 (UTC)

  I understand that. I don't think this is a good criterion to use. I don't want to wade deeper into the AI on Wikipedia debate again. However, I will say that trying to divine whether something is unreviewed AI output based on how it is written is primarily an aesthetic issue. Text is text. If this was about content rather than aesthetics, then the only criterion in G15 would be "communication intended for the user" and there wouldn't be any reference to subjective signs in G15. lethargilistic (talk) 14:58, 11 August 2025 (UTC)

• Oppose per others as markdown is common elsewhere and can be fixed with a simple find-and-replace. 123957a (talk) 07:30, 11 August 2025 (UTC)
• Weak support but only if both reference url format and content (*/**) markup is used throughout. Lots of the opposes point out it is common, but before LLM Slop I never saw a single AfC submission using it and certainly not the link format. KylieTastic (talk) 10:18, 11 August 2025 (UTC)
• Strong Oppose Use of Markdown is not a reliable enough indicator of AI or LLM usage (or generated slop) any more than any other markup. In fact, outside working in Microsoft Word, most text-based editors allow option to export to Markdown including Google Docs. It is not unreasonable to believe that an editor working on a draft article may have markdown markup in their text. A CSD rationale based would be a hostile overreach of a genuine problem that the template may be trying to solve. WeWake (talk) 20:02, 13 August 2025 (UTC)
• Support, if qualified as internally consistent and widespread throughout the page. As with the other points already codified in G15, a single instance of poor formatting or markdown is not alone indicative of AI use, but if all section headers and citations incorporate identical markdown in conjunction with other tells, unreviewed LLM output is the most logical explanation. Some pages I've seen tagged for G15 (and sometimes G11) incorporate markdown with impeccable consistency, and I believe that the preview button and self-reversion should largely prevent users from making the same formatting mistakes so consistently (i.e., there would be fewer mistakes, and of varying nature). I believe this qualifier could be worded objectively to filter out honest mistakes, and in edge cases (well, all cases), the reviewing admin should exercise their best judgement.

Additionally, I would support a grace period as with A1 and A3 (i.e., do not tag less than 15 minutes after creation), as it would give time for those actively working on a page to bring it up to par, while those lazily copy-pasting LLM output would probably not be bothered to make a second pass (neither fixing obvious formatting mistakes nor contesting the speedy deletion). And as a side note, in my experience facilitating many IRL events, newbies mostly gravitate towards visual editor and in turn are less likely to encounter wikitext while they're still learning the ropes – so I'm not overly concerned about a badly-tagged page driving away a newcomer acting in good faith. ^Complex/_Rational 22:21, 17 August 2025 (UTC)

• Support subset: "Faulty wikitext that mixes in Markdown syntax" (in the same revision, not counting partial or faulty manual attempts to convert Markdown to wikitext) per my comment below and per WP:MARKDOWN (I wrote that, just to declare for transparency). This is strong evidence that the user told a chatbot to create a Wikipedia article. It's stronger than the existing subcriteria "Nonsensical citations" and "Implausible non-existent references". —Alalch E. 15:54, 19 August 2025 (UTC)

  No. I think the efforts needed to distinguish "faulty manual attempts" from real LLM usage is nontrival and involves large amounts of subjective evaluation. MilkyDefer 06:44, 21 August 2025 (UTC)

  @MilkyDefer I wrote in the same revision, not counting partial or faulty manual attempts to convert Markdown to wikitext meaning to say that those "faulty manual attempts" would occur in a subsequent revision and if they were to occur, then G15 would not apply (the opposite of would apply), so we actually agree. If the first revision is all-Markdown (because, let's say, the creator thought Wikipedia supports Markdown and created their article in Notepad which now supports Markdown) and in a subsequent revision there is any attempt to fix that (causing a faulty mix of wikitext and Markdown syntax), then G15 would not apply. —Alalch E. 22:16, 23 August 2025 (UTC)

  But MediaWiki has a preview button. This makes the mixed markup in the same revision a possible outcome of benign faulty attempt to fix. Perhaps someone uses the preview button and realizes that the layout is screwed, tries their best to fix but gives up and publishes whatever as-is. MilkyDefer 03:10, 24 August 2025 (UTC)

  That's unlikely enough for speedy deletion to be appropriate. —Alalch E. 07:13, 24 August 2025 (UTC)

oppose per Sohom. markdown is pretty common, some newbies may think wikipedia uses it. drinks or coffee _{or prime} *GET OUT* 04:40, 30 August 2025 (UTC)

Discussion (Markdown)

• I think that if this criterion is added, it should come with a qualifier stating that Markdown is not an unambiguous sign of LLM usage; instead, it commonly appears alongside other signs, like the weird "oaicite" things or nonexistent references. As has been pointed out above, newbies (and experienced editors) may accidentally use Markdown on a page, so we should not use Markdown as the only reason for speedy deletion. This is my suggestion to add:

  Since humans may accidentally use Markdown instead of wikitext, this should not be considered as a definitive sign of LLM-generated content. It may be treated as one if it appears alongside other, more definitive criteria.

  It should still probably be included (in my opinion) because it is much easier to spot on pages than some of the other signs of LLM writing; in fact, Markdown has an edit filter already. We could always treat Markdown as a preemptive sign of LLM-generated content even if it isn't explicitly mentioned in the text of the CSD. SuperPianoMan9167 (talk) 23:16, 4 August 2025 (UTC)

  We should avoid mentioning anything in this criterion that is not an unambiguous sign of unreviewed-LLM use. Leaving markdown in when reviewing output (whether LLM-generated or otherwise) is something a new editor unfamiliar with Wikipedia is liable to do, it's a reliable tell for an inexperienced editor, but the (ostensible) purpose of this criteria is to get rid of content that people suspect to be LLM-generated, not to bite newbies. Thryduulf (talk) 00:30, 5 August 2025 (UTC)

• Are there any examples of pages (extant to deleted) that would qualify with this criterion? (If there are very few then there's no need to clutter up G15.) Graeme Bartlett (talk) 11:41, 5 August 2025 (UTC)

  Virtually every page that I've nominated for G15 because of the ":contentReference[oaicite:1]{index=1}" things also has Markdown formatting in it. Most of them have since been deleted; some of the ones currently awaiting deletion that have been recently deleted are Draft:Arud Village and Draft:Tutor Koraphat Lamnoi. (Explicit deleted those pages as I was writing this comment.) SuperPianoMan9167 (talk) 11:46, 5 August 2025 (UTC)

  It has just been deleted, but I just tagged User:Akasha Official for deletion under G15, remembered this conversation, then changed it to G11. It was very clearly LLM-generated, with vague promotional language ("He is married and continues to inspire young individuals through his lectures, digital content, and public speaking engagements."), excess bolding and bolded bulleted lists, and random markdown, however there were no citations. Sarsenet•he/they•(talk) 12:02, 5 August 2025 (UTC)

  If the pages are getting deleted already then adding this would be redundant at best and increase the false positives at worst. Thryduulf (talk) 12:33, 5 August 2025 (UTC)

  contentReference[oaicite:1]{index=1} is highly specific for chatgpt, and is a kind of bug there. So it proves LLM generation. But then the question will be, has it been edited by a human? Graeme Bartlett (talk) 23:02, 5 August 2025 (UTC)

• G15 is still new and we need some time to evaluate its effectiveness. If we find that G15's narrow criteria are failing to catch obviously LLM-generated pages that contain Markdown syntax, then we should revisit this proposal to include Markdown in the criteria. — Newslinger talk 14:16, 9 August 2025 (UTC)
• Theoretically, it is possible to define a specific scenario that would fit the essence of the criterion like "Faulty wikitext with mixed-in Markdown syntax, enclosed in a Markdown fenced code block" (```wikitext/```markdown/``` near the top and/or ``` near the bottom). As seen in:
  • Special:Permalink/1290687796: Wikitext markup for boldface, bullets, and headings but Markdown syntax for external links, ``` is present at the bottom.
  • Special:Permalink/1297827841: Wikitext markup for some of the boldface, bullets, headings, and links, but Markdown for the rest of boldface (in the "Selected Projects" section), and ```wikitext is present at the top.
  • Special:Permalink/1293085602: Wikitext markup for the infobox, boldface, bullets, and headings but Markdown syntax for external links, and ```markdown is present at the top and ``` at the bottom.
  The presence of backticks is actually similar to "Communication intended for the user", but in the scenario I am describing here, the user has almost certainly removed preceding text similar to "Here is the Wikimedia code for Nishith Goswami:" (permalink) or "--- Here is the **Wikipedia markup (wiki code)** for the article on *The Retail Song Cycle*: (permalink)—or the entire chatbot's output was a fenced code block, as a chat reply to something like "convert to Wikipedia source code" or "yes, please"—as an answer to "If you're interested in actually submitting this to Wikipedia, I can help convert it into proper wiki format ..." (permalink).
  I think that the backticks which are clearly not human-inputted (why would they be?) + broken formatting, which shows to the creator that the wikitext outputting or conversion to wikitext was not performed without error (and simultaneously shows the user which markup features work and which don't enabling them somewhat to fix the content), is obvious enough that the broken formatting would have been removed/fixed by any reasonable human review.
  (Some pages have backticks and correct wikitext, showing a more advanced use of AI: Draft:Serafim FUTURES®, Draft:Ensis AI)
  Ping Sohom_Datta for an opinion.—Alalch E. 20:15, 14 August 2025 (UTC)

  (I forgot to respond to this) I think starting with ```wikitext or ```markdown could potentially be considered "communication intended for the user", but it's less ambiguous than the explicit that the text itself. The three backticks represent code blocks in markdown and I can see folks doing that pattern as well as part of their notes. Sohom (talk) 22:44, 23 August 2025 (UTC)

  Thanks for the reply. I can't see people doing that as part of their notes. How would that happen? —Alalch E. 23:01, 23 August 2025 (UTC)

The discussion above is closed. Please do not modify it. Subsequent comments should be made on the appropriate discussion page. No further edits should be made to this discussion.

G15 clarification #2

Would "communication intended for the user" apply to edit summaries? Draft:Chip Shelton was created with the following edit summary:

Added neutral, Wikipedia-style lead summarizing Chip Shelton’s biography, notability, and key achievements, maintaining encyclopedic tone and reliable sourcing.

I would normally expect a summary for draft creation like "Draft creation using the WP:Article wizard", or else an automatic edit summary like "Created page with [content]". As I was writing this comment, the author (EARockstar, whose account is only three hours old) expanded the article with this edit summary:

Replaced incomplete text with full Wikipedia-formatted biography of Chip Shelton, including lead, career, discography, innovations, outreach, and references.

These edit summaries point out the obvious (this is Wikipedia, so of course a Wikipedia article would have a Wikipedia-style lead) in a way that suggests the draft was generated by a LLM, especially because the draft originally had Markdown in it (which is likely not going to be an explicit part of G15 due to the strong opposition in the RFC above). Is this page eligible for G15? I'll avoid tagging the page with {{db-g15}} for now in case the draft is not eligible. SuperPianoMan9167 (talk) 21:17, 9 August 2025 (UTC)

My view is that the criterion as currently written does cover this case - nowhere does it say that the text of the article itself must contain the LLM instruction. "page that exhibits one or more of the following signs which indicate that the page could only plausibly have been generated by large language models ... Communication intended for the user" reads to me as inclusive of directly relevant text such as the page creation edit summary. If that's not our intention we should probably adjust the criterion to read something like "pages that contain one or more..." Sam Walton (talk) 21:43, 9 August 2025 (UTC)

I went ahead and nominated the page; if an admin deletes it, we should probably modify the criterion to make it clear that the signs can appear in edit summaries. SuperPianoMan9167 (talk) 21:54, 9 August 2025 (UTC)

That's kind of backwards - this policy describes what kind of pages the community permits individual admins to delete on their own recognizance, not what admins think they shouldn't have to bother with a weeks-long consensus-based deletion discussion for. There have been admins who have batch deleted entire categories of speedy-deletion-tagged pages at once without, so far as anyone else could tell, looking at them; taken to its logical extreme, your position would have made all of those tags ok because an admin pushed the delete button.

Similarly, I don't think it wise to rely on the precise wording of the criterion where that detail hasn't been debated yet. I don't see any indication in the rfc above that those supporting considered whether anything not in the current revision, let alone not in the page itself, could support a speedy. The closest, I think, was my own comment in #Discussion (LLM) about looking at drafts as of their first submissions; but even there I was trying to determine whether they could have been speedied then and saved further review, not whether they could still be speedied now.

I guess the answer to whether we should consider edit summaries will be the same as for Jumpytoo's still-unresolved question towards the end of #G15 is not self-evident about whether a page is still speedyable if the definitive sign is removed but the rest of the page left unchanged. Or, for that matter, whether we could speedy an LLM-looking page without one of these signs if every single one of the editor's other edits are unquestionably LLM-generated. —Cryptic 22:17, 9 August 2025 (UTC)

Here's a couple of statements in response:

1. I agree with you that "did an admin delete it?" is a pretty poor reason for adding to a CSD; that should be determined by discussion instead. Like you said, the fact that an admin deleted a page does not necessarily mean it meets the criteria. That's why I opened this discussion; if there is consensus for modifying the criterion, that'll be what modifies it, not the fact that an admin deleted it. Thank you for pointing this out.
2. The author's edit summary of "Replaced lead with concise, fully sourced version in my own words." (emphasis added) is a pretty clear admission that they did in fact use LLMs to make the draft. They are still probably using LLMs because they made another edit that replaced most of the formatting with Markdown.
3. I agree with other editors in the aforementioned section #G15 is not self-evident that the problem with these pages is not due to the signs of LLM content themselves; the problem is "unreviewed LLM-generated content". Just removing the indicators does not fix the problem.

SuperPianoMan9167 (talk) 22:54, 9 August 2025 (UTC)

The author of the aforementioned draft has actively made changes to the draft and worked on improving it, so it is now ineligible for G15 as it is not LLM-generated content without reasonable human review. It still probably is LLM-generated content, but it can't be speedily deleted anymore. SuperPianoMan9167 (talk) 08:38, 10 August 2025 (UTC)

The lead of the policy explicitly states A page is eligible for speedy deletion only if all of its history is also eligible. nothing in the discussion about G15 resulted in a consensus to change that or to apply it differently. We want editors to improve their drafts. If they start with something that is LLM-generated and then edit it to fix any problems it contains then that is a good thing.

Edit summaries should not, imo, be used as an indicator of anything more than a pointer to check whether the content of the article meets a speedy deletion criterion, and should never be determinative on its own. Thryduulf (talk) 11:41, 10 August 2025 (UTC)

This seems like the editor asked the LLM to summarize the edit, and its possible the editor reviewed the LLM changes before asking it to summarize it for the edit summary. So I would not think that is enough for G15.

Perhaps if the edit summary was something like Sure! I've modified the article as requested to add an neutral, Wikipedia-style lead summarizing Chip Shelton’s biography, notability, and key achievements, maintaining encyclopedic tone and reliable sourcing, up to the information known in my last training update. I would say is eligible. Jumpytoo Talk 05:01, 11 August 2025 (UTC)

I'll affirm that using LLMs to summarize human-made edits is, weirdly, a real phenomenon. It is not helpful and I wish people wouldn't, but I wouldn't consider an LLM-generated edit summary to be de facto proof that the edit itself consists of unreviewed AI outpit. ~ L 🌸 (talk) 07:55, 6 September 2025 (UTC)

"Should" notify the page creator?

The beginning lines of the SK policy contains, two obligations for users nominating a page for speedy deletion:

• Should specify which criterion/criteria the page meets
• Should notify the page creator and any major contributors

Therefore

1. Is it against the policy to either nominate or delete a page for speedy deletion without letting the creator know?
2. Is the specification of the criterion in the deletion log also mandatory?

Xpander (talk) 15:39, 10 August 2025 (UTC)

Generally you should let the creator and any major contributors know, but there are exceptions. For example if the editor is banned or hasn't edited in years then there is no benefit to doing that. The deletion log needs to make it clear why the page was deleted, in the case of speedy deletion that means it is normally best to include a reference to the criterion but it isn't mandatory. For example if the page was nominated at XfD and the consensus there was to speedily delete, it's fine (arguably better) to link to that discussion rather than/in addition to the speedy criterion. Thryduulf (talk) 16:02, 10 August 2025 (UTC)

What may be the repercussions for editors who nominate articles for speedy deletion without notifying the article creators? Significa liberdade (she/her) (talk) 16:28, 10 August 2025 (UTC)

I believe "should" is inadequate and should be "must." Having your hard work deleted feels like a slap in the face. At the very least, we owe the page creator an explanation for the tagging. Better to give them an advisory welcome as well, such as "first article." Not notifying them is not assuming good faith and is bitey. Believe me, a welcome can relieve the sting and make the difference in retaining/helping new editors. -- Deepfriedokra (talk) 16:43, 10 August 2025 (UTC)

It's not all hard good-faith work though is it? I mainly delete pages under G2, G3, G5, and G10, maybe G6 and G11 (and often IAR). No notification is usually appropriate (especially G3 and G10). Instead of making it mandatory and carving up the exceptions in the introduction, it's probably right to just say what is normally appropriate. -- zzuuzz ^(talk) 16:58, 10 August 2025 (UTC)

You should never be speedily deleting anything under IAR. The criteria list the only times when deletion without discussion is permitted, other speedy deletion is inherently controversial. IAR is only to be used when the action uncontroversially improves the encyclopaedia and thus cannot be used to speedy delete things that are not covered by one or more of the criteria. Thryduulf (talk) 17:06, 10 August 2025 (UTC)

Any page that should be oversighted because it contains inappropriate personal information should be immediately speedily deleted. Yes, the admin should also request oversight, but any admin who worries whether the page is covered by the criteria instead of IAR deleting is doing it wrong. —Kusma (talk) 17:13, 10 August 2025 (UTC)

Exactly. Sure, your typical newbie's A7 or G12 should come with notification, but a general "must" isn't the right way forward. —Kusma (talk) 17:07, 10 August 2025 (UTC)

Those new editors who produce WP:G11 and WP:U5 content have no idea why we are deleting their hard work. They are acting in good faith, but don't know our rules. We cannot expect them to follow the rules if we do not explain the rules to them. Notifying them of speedy deletion tagging is the minimum education we can provide. -- Deepfriedokra (talk) 17:35, 10 August 2025 (UTC)

But there are always exceptions. Take this spambot for example. A good faith spambot? Hard work? -- zzuuzz ^(talk) 17:39, 10 August 2025 (UTC)

Sometimes those we think are spambots really are not. -- Deepfriedokra (talk) 17:42, 10 August 2025 (UTC)

And I am referring to real new users and not spambots -- Deepfriedokra (talk) 17:43, 10 August 2025 (UTC)

Thank You, I am using this platform to help the culture and community. Some misinformation was put on here intentionally about Word Up Magazine. Now it’s circulating as truth. Please don’t allow this. How can we correct it? Marcus G Blassingame (talk) 05:32, 16 August 2025 (UTC)

Regardless of whether we're deleting someone's hard work, what's the harm of notifying them that the page is being deleted? Significa liberdade (she/her) (talk) 23:40, 10 August 2025 (UTC)

In the case of this spambot, the harm extends to a total waste of my time, or the nominator's time, plus an extra talk page that no one is going to read. In the case of someone like User:I eat ass and pussy, it's the creation of an offensive title, wasting someone's time for an LTA, and encouraging them by providing recognition (plus the no one reading and possibly eventual deletion). In the case of User:Azhar Morgan, it's around 200 notifications (which no one is going to read because the account is globally locked). In the case of the creator of Poopy did a diarrhea shiatted in her pants, it's something everyone already knows, plus several things already mentioned. I'm not saying that notification shouldn't be provided for the people Deepfriedokra and others describe and normally think about when it comes to deletion - I'm one of the people who even thinks that retired users should normally be notified. But if it were mandated it would be ridiculous in several situations, and the exceptions list would never be satisfactory. "Normally", yes; "Should", yes; "Must", nah. -- zzuuzz ^(talk) 06:17, 11 August 2025 (UTC)

If the nominator is using a tool such as Twinkle, the time it takes to notify the page creator and major contributors is basically null. The deleting admin doesn't have any additional work to do. I'm not sure how notifying editors results in a significant amount of wasted time. Significa liberdade (she/her) (talk) 18:56, 11 August 2025 (UTC)

@Kusma, Just to be clear, are you suggesting (I have in mind 'A7' in particular) that in case of new users, they should be notified but not for older users? Xpander (talk) 19:52, 10 August 2025 (UTC)

We should usually notify (with a few exceptions that I would not like to legislate). It is especially important for newbie articles, and A7 and G12 are typical newbie mistakes so I would hope that they get a notification so they have someone to ask about it. —Kusma (talk) 19:56, 10 August 2025 (UTC)

The overarching rationale here, being that, if the creator is not notified, how should they find out? Through chance? Xpander (talk) 19:55, 10 August 2025 (UTC)

Especially for new users, lack of a deletion notification can cause excessive confusion and even recreation of the same articles over and over as they either don't know why their work has been deleted or they don't even know it has been deleted. Significa liberdade (she/her) (talk) 23:42, 10 August 2025 (UTC)

Who we should be notifying is the major editor or editors of the deleted content. That's very often, but not always, the same as the page creator. If you're only relying on Twinkle to do the right thing, or trying to effectively mandate the use of Twinkle or similar tools by limiting it to the creator (who may have just made a 20-byte redirect, not the 10,000-byte article it was expanded to), you're doing it wrong. —Cryptic 23:48, 10 August 2025 (UTC)

I guess the question then is whether it's "against the policy to either nominate or delete a page for speedy deletion without letting the creator [and significant contributors] know". Significa liberdade (she/her) (talk) 23:55, 10 August 2025 (UTC)

And the answer is "no." However, your rationale for notification is sound. I mostly delete spam drafts and user pages, so for me notification is essential, and TWINKLE is usually (not always) up to the task of notification. The trick is to be smarter than your tools. -- Deepfriedokra (talk) 00:09, 11 August 2025 (UTC)

Most of my deletion-related activity is related to redirects. It's not too uncommon for the creator to be the only significant contributor and to have not edited in over 5 years, to be a bot and/or be banned. In those cases notifications really don't help anybody. However there are plenty of other cases (almost certainly the majority) where notifications absolutely do add value and I do leave those notifications. Thryduulf (talk) 02:03, 11 August 2025 (UTC)

Even if the notification is worthless, I don't think it's harmful. If you're using a tool such as Twinkle, it also shouldn't waste any time. So the question is why you shouldn't inform them. So what if it goes to a retiree or bot or inactive user? Significa liberdade (she/her) (talk) 18:25, 11 August 2025 (UTC)

I don't use Twinkle or similar tools, so it does take time (not much, but not-zero) and pointless notifications could be taken as spamming. Thryduulf (talk) 19:12, 11 August 2025 (UTC)

I'm going to express the minority opinion that you shouldn't need to notify anyone of anything; if they still care about the page it should be on their watchlist. * Pppery * _{it has begun...} 04:24, 11 August 2025 (UTC)

It takes a while for people to get a hold of their watchlist, or other monitoring tools, even then they might not check it that often. In any case what's the harm in a simple notifying? Provided that it's not clear vandalism. Xpander (talk) 08:30, 11 August 2025 (UTC)

And the great thing about TWINKLE is it notifies with the same mouse click that tagged the page. -- Deepfriedokra (talk) 08:42, 11 August 2025 (UTC)

Beyond issues regarding the watchlist, this also assumes editors understand what's happening. New editors might create an article, then have it speedy deleted within 24 hours, in which case they might have no idea why their article was deleted. At least with basic notifications, they get a blurb explaining the rationale. Significa liberdade (she/her) (talk) 18:23, 11 August 2025 (UTC)

I can't imagine them not seeing the deletion summary itself. Since the notifications are impersonal form-letters based on the deletion summary anyway, they don't say anything it doesn't. * Pppery * _{it has begun...} 18:24, 11 August 2025 (UTC)

Looking at one article I just speedy deleted, the deletion summary says, "Multiple reasons: speedy deletion criteria U5, G11". Conversely, the deletion notification provides a basic summary of each deletion rationale. Even if it's templated, that's a lot more helpful to newbies than just seeing a link, especially given that many people do not follow the link. Significa liberdade (she/her) (talk) 18:29, 11 August 2025 (UTC)

Additionally, the notification provides the user time to actually go write a contestation. Significa liberdade (she/her) (talk) 18:30, 11 August 2025 (UTC)

I agree with Significa liberdade -- Deepfriedokra (talk) 18:48, 11 August 2025 (UTC)

In general, I think "should" is fine for admin-facing policy that has reasonable exceptions. Admins are people that generally do what they should, and those that routinely don't will lose the bit. Firefangledfeathers (talk / contribs) 15:11, 11 August 2025 (UTC)

What about non-admins who regularly nominate articles for speedy deletion (including outside of vandalism)? Significa liberdade (she/her) (talk) 18:20, 11 August 2025 (UTC)

Oof! I realized this was about nomination and not deletion and then I went away and the realization slipped into another dimension. I'm more on the fence about must v. should for nominations. Firefangledfeathers (talk / contribs) 18:28, 11 August 2025 (UTC)

Rather than repeat myself, I'll just link to User:Ritchie333/Don't template the retirees. Ritchie333 ^{(talk) (cont)} 09:56, 11 August 2025 (UTC)

"Should" is fine. It'll usually be appropriate. It occasionally won't. We don't want to put in hard blanket rules. Stifle (talk) 09:16, 12 August 2025 (UTC)

In addition to the examples raised above, there's no benefit to littering sock talkpages with G5 notifications. CMD (talk) 09:57, 12 August 2025 (UTC)

I concur with others above that "should" means, well, "should". There's good reason for it not to say "must", and there's good reason for it not to say nothing. Notify unless there's a common-sense reason not to notify. If someone doesn't notify, a polite reminder is appropriate, although with an experienced nominator framing it as a query might be better, since sometimes there can be rather obscure but still very good reasons not to notify, like "The user told me 5 years ago not to bother notifying them". -- Tamzin^{[cetacean needed]} (they|xe|🤷) 16:46, 15 August 2025 (UTC)

+1. asilvering (talk) 16:51, 15 August 2025 (UTC)

+2. Tamzin nails it. Also, if you think that I must notify some abusive sockmaster that I've G10/G5ed their perpetual death threat creations or attack pages, well then you can have all of my tools back. It shouldn't be easier for trolls to troll than for admins to disappear their garbage into the ether.-- Ponyo^{bons mots} 17:22, 15 August 2025 (UTC)

What would you recommend as the next course of action in the following scenario? An editor has an edit count over 20k and has been on Wikipedia for over 5 years. They regularly nominate new articles for speedy deletion in non-vandalism areas (e.g., A7, G11, U5), but they never inform the article creator. The editor has been notified more than once about informing article creators, but they never respond to messages and continue not to inform article creators. Significa liberdade (she/her) (talk) 19:14, 15 August 2025 (UTC)

@Significa liberdade: he did actually respond to my second inquiry. He does notify those he does not deem unworthy. To paraphrase, "he ain't gotta; he ain't gonna." As he is correct about notification not being a requirement, I just don't worry about it anymore. -- Deepfriedokra (talk) 19:20, 15 August 2025 (UTC)

AN/I. Or, for cowboy-admin types, a final warning for disruptive editing (WP:DISRUPTSIGNS ##4-5) and then a tempblock if it happens again. -- Tamzin^{[cetacean needed]} (they|xe|🤷) 19:21, 15 August 2025 (UTC)

Nah. Too much like work. -- Deepfriedokra (talk) 19:23, 15 August 2025 (UTC)

• I can think of scenarios where it would be appropriate not to notify, such as LTAs who tend to harass anyone who crosses their path and someone who has explicitly asked not to receive such notifications. So, I think "should" is the correct terminology—under normal circumstances, one certainly should give such a notification, and that should be the standard practice, but there are reasons to legitimately deviate from it. ("I don't feel like it today" is not such a reason.) Normally, policy has taken the approach of treating editors as rational people who can be trusted to exercise a reasonable amount of discretion and to recognize when an exceptional situation arises, and I think the vast majority of the time, that is the approach we should (though, I suppose, not must) take. If someone is clearly abusing that discretion, in this case if they were rarely or never notifying page creators, we can use kind of an "ask, tell, make" sequence to address that. Seraphimblade ^{Talk to me} 06:55, 16 August 2025 (UTC)
• Many times, notifying is not appropriate. Changing to "must" would be a giant bureaucratic headache, as each time, you would find some wikilawyer contrarian dragging people to ani for something that is obvious to the rest of the world, wasting time that would be better spent actually improving the encyclopedia. If I don't notify someone about a speedy delete, there is a valid reason. You are still free to notify them if you disagree. Dennis Brown - 2¢ 06:16, 20 August 2025 (UTC)
• Noting related discussion HERE.-- Deepfriedokra (talk) 01:44, 23 August 2025 (UTC)
• I think a huge point here is that "should" means different things to different people. "Should" could mean "do it unless you have a valid reason not to." It could also mean "this is best practice but do what you want." I tend to fall into understanding it as the former, especially given that we have policies that state something along the lines of "it is recommended to do X," which I feel counts more toward "best practice but do whatever." Significa liberdade (she/her) (talk) 15:48, 26 August 2025 (UTC)

  As an example of "recommended" language, AfD guidelines state, While not required, it is generally considered courteous to notify the good-faith creator and any main contributors of the articles that you are nominating for deletion. If we're taking "should" to mean "it's generally courteous but not required," we should say that. Significa liberdade (she/her) (talk) 17:25, 26 August 2025 (UTC)

  @Significa liberdade: According to rfc:2119 (which some people are aware of but many are not), the word "should" means that there may exist valid reasons in particular circumstances to ignore a particular item, but the full implications must be understood and carefully weighed before choosing a different course. This is much closer to "do it unless you have a valid reason not to" than it is to "this is best practice but do what you want". --Redrose64 🌹 (talk) 07:16, 27 August 2025 (UTC)

  Thank you! Significa liberdade (she/her) (talk) 14:39, 27 August 2025 (UTC)

RfC (notifications)

The following discussion is an archived record of a request for comment. Please do not modify it. No further edits should be made to this discussion. A summary of the conclusions reached follows.

Clear consensus for the status quo. Extraordinary Writ (talk) 03:24, 22 September 2025 (UTC)

---

I've been following this discussion for a while, and seeing how someone was reported to ANI for not notifying page creators, I think it's time for an RfC

• Option 1: Page creators must be notified when their article gets tagged for speedy deletion.
• Option 2: Page creators must be notified when their article gets tagged for speedy deletion, excepting obvious vandalism, attack pages, or pages otherwise created in bad faith.
• Option 3: Page creators should be notified at the discretion of the nominator.

(this is my first RfC so please let me know if I'm doing something wrong.) Children Will Listen (🐄 talk, 🫘 contribs) 02:32, 23 August 2025 (UTC)

• Option two as first choice; option three a second choice. There are very good arguments made above for not notifying everybody. However, in some instances where good faith must be assumed, it is bitey not to notify them. -- Deepfriedokra (talk) 02:44, 23 August 2025 (UTC)
• Option 3 per my comment above. * Pppery * _{it has begun...} 03:06, 23 August 2025 (UTC)
• This is entirely unnecessary. We already have a clear consensus above that the status quo is fine. And all three options are worse than the status quo in that they only mention the page creator, and further normalize the terrible practice of just clicking on Twinkle without investigating who the major contributors are; and "at the discretion of the nominator" is also weaker than the status quo. —Cryptic 03:11, 23 August 2025 (UTC)
• None of the above. The current wording is fine. -- Whpq (talk) 03:18, 23 August 2025 (UTC)
• None of the above. (Summoned by bot) Having read the above discussion, The current wording is fine to me.Pincrete (talk) 04:30, 23 August 2025 (UTC)
• Status quo per Cryptic (major contributors other than the page creator, if any, should also be notified), Option 3 as second choice per above discussion (another example would be not notifying a deceased wikipedian to not clutter up their talk page: a "must" wording is too restrictive in this case). OutsideNormality (talk) 04:44, 23 August 2025 (UTC)
• Status quo. If you want to legislate exceptions, there's at least a dozen more that need to be listed, but why? The discussion above shows that the current wording communicates what's intended: Do it, unless there's a good reason not to. -- Tamzin^{[cetacean needed]} (they|xe|🤷) 04:48, 23 August 2025 (UTC)
• None of the above. The current status is fine - and succinct: "Users nominating a page for speedy deletion should specify which criterion/criteria the page meets, and should notify the page creator and any major contributors." Un point, c'est tout.-The Gnome (talk) 05:43, 23 August 2025 (UTC)
• None of the above. The status quo is superior to all the options presented. Thryduulf (talk) 10:12, 23 August 2025 (UTC)
• Oppose 1 and 2 The only times when a page-creator deletion notification to User talk:AnomieBOT has been useful is when it's from someone uninformed enough to nominate only an ASCII hyphen-minus redirect title without also nominating the corresponding en-dashed title. Anomie⚔ 14:12, 23 August 2025 (UTC)
• None of the above - I think the wording is fine as it is. Cheers, SunloungerFrog (talk) 15:06, 23 August 2025 (UTC)
• Status quo not seeing a compelling reason for a change here and I agree with Cryptic's concerns. -- LWG ^talk 15:53, 23 August 2025 (UTC)
• Status quo per above. The existing wording is pretty clear actually and means what it should mean.—Alalch E. 23:51, 23 August 2025 (UTC)
• Status quo - This is a solution looking for a problem. Seems an incredible waste of time trying to push a change against a well established and very clear consensus. Dennis Brown - 2¢ 01:34, 24 August 2025 (UTC)

  The problem to me, as I've stated above, is that it isn't clear what "should" means. Does it mean, "Do so unless you have a valid reason not to?" or does it mean "it's considered good good practice"? If it's the latter, why don't we just say that? If it's the former, I feel there should be some for of community repercussions for following guidelines. Significa liberdade (she/her) (talk) 17:30, 26 August 2025 (UTC)

• Weather Alert It's snowing in August! -The Gnome (talk) 10:54, 24 August 2025 (UTC)
  • Off topic, but I've often wondered what those in the southern hemisphere make of comments like this? Thryduulf (talk) 13:32, 24 August 2025 (UTC)
    • It must be a bit like us in New England hearing the expression "coal to Newcastle", since the towns named Newcastle in these states are certainly not centers of coal production. The Blade of the Northern Lights (話して下さい) 18:31, 24 August 2025 (UTC)
    • For southern hemisphere contributors, their weather cycle can be used as a refutation of the snowball clause. -The Gnome (talk) 21:22, 25 August 2025 (UTC)

The discussion above is closed. Please do not modify it. Subsequent comments should be made on the appropriate discussion page. No further edits should be made to this discussion.

RfC: Including emojis in G15

Should the following criterion be added to G15? ~/Bunnypranav:<ping> 12:22, 27 August 2025 (UTC)

• Emojis: This may include usage of Emojis with no encyclopedic value, only if used in the mainspace and draftspace page at the beginning of paragraphs or in place of bullet points.

---

Survey (Emojis)

• Support as proposer. As also noted at Wikipedia:Signs of AI writing#Emoji, LLMs are notorious for leaving emojis in their output. Such articles can be safely deleted because human editors will most probably not include emojis in their writing, unlike the possibility with markdown (older RfC). ~/Bunnypranav:<ping> 12:22, 27 August 2025 (UTC)
• Support, with caveat. I agree that this is highly indicative of AI and that it is very unlikely to be used by a normal human editor. I acknowledge that it is possible that a human editor reviewing LLM output would leave emojis in articles, but find it unlikely. There is a bit of a complication here in that this is a G criterion and not an A one. I wouldn't approve of someone G15'ing a userpage for this reason, for example - someone who used AI to create a quick userpage shouldn't have it deleted as "unreviewed by a human" if they used emoji bullet points to describe their editing interests. In drafts or articles, however, I agree that this demonstrates lack of human review and is thus in line with the spirit of G15. -- asilvering (talk) 18:48, 27 August 2025 (UTC)

  If we take the relevant content to be content intended to be published as an article such as a draft article, a userspace draft, or a WP:FAKEARTICLE, or article content published in entirely wrong namespace, such as a talk space (see example [no emojis, but doesn't matter]: Special:PermanentLink/1306518478#North Indian and south indian bhatt Brahman linkage, and their direct relationship with bhatraju of south andhra telengana karnataka bhatrajus), then a general criterion is suitable. —Alalch E. 20:41, 27 August 2025 (UTC)

• Support, as I never saw humans use emojis on Wikipedia like the chatbots do.—Alalch E. 20:43, 27 August 2025 (UTC)

  I have seen GreenLipstickLesbian's analysis below, but "used in the mainspace and draftspace page at the beginning of paragraphs or in place of bullet points" is sufficiently indicative and discriminate. —Alalch E. 14:08, 2 September 2025 (UTC)

  If you've never seen humans use emojis in place of bullet points, then I bet you aren't old enough to remember Zapf Dingbats. WhatamIdoing (talk) 20:44, 6 September 2025 (UTC)

• Oppose per my comment and example below; not an unambiguous LLM tell. In fact, I'm pretty sure humans are more likely to use emojis than markdown, and that proposal is getting pretty resoundingly vetoed. (Just to talk about biases, I think the fact that Wikipedia editors tend to be very techy makes us think markdown is more common than it is. It's certainly not more common than emojis, anyway.) GreenLipstickLesbian💌🦋 21:50, 27 August 2025 (UTC)
• Support, provided that the offending page is in articlespace or is intended by the creator to be an article at some point (no matter how implausible), such as having {{Userspace draft}} or {{AfC submission}} (not {{User sandbox}}), being in draftspace or being submitted to AfC; we don't need to police users using emoji on their user page. Also, the criteria should limit the emojis to only being in headings, lists or heading or list-like constructs, only being for (ostensibly) aesthetic purposes (explicitly excluding all flags, ❌/✔, male/female, astrological and mythological symbols (e.g. ♏), and emojis with defined or implied legends), not being all the same (explicitly excluding new editors attempting to emulate lists using bullet-like emojis), and most of them being in emoji style (i.e. not having a unicode variation selector or CSS forcing text) per Special:GoToComment/c-GreenLipstickLesbian-20250827214700-Bunnypranav-20250827141900. OutsideNormality (talk) 02:03, 28 August 2025 (UTC)
• Oppose At least not without further workshopping. Examples given below have demonstrated that LLMs are not unique in their use of emojis. I think OutsideNormality provides a good starting point for narrowing down the proposed criterion. I am also worried that this indicator will soon be obsolete. According to OpenAI, GPT-5 now uses "fewer unnecessary emojis", and I expect that other LLM makers will follow suit. Ca ^{talk to me!} 07:24, 28 August 2025 (UTC)
• Conditional Support Per GreenLipstickLesbian, emojis are even more likely to be used by novice human editors than markdown, the latter of which appears en route to rejection as a G15 criterion. Even while knowing that they are unprofessional for article text, I use them in plenty of edit summaries as tone indicators. Accordingly, I only support this criterion if limited to use at the beginning of paragraphs or in place of bullet points. ViridianPenguin🐧 (💬) 11:30, 28 August 2025 (UTC)

  Note that the proposed criteria only applies if emojis are used "in a mainspace or draftspace page at the beginning of paragraphs or in place of bullet points." No offense or rebuke intended to ViridianPenguin. I would insert a smiley with cold sweat running off brow emoji here if I remembered how to do so.--FeralOink (talk) 08:01, 9 September 2025 (UTC)--FeralOink (talk) 08:01, 9 September 2025 (UTC)

  Sorry, nevermind! I see that the criteria was amended after you made your comment.--FeralOink (talk) 08:10, 9 September 2025 (UTC)

• Support per various comments above - support conditional on it only applying to mainspace and draftspace, and only when the emojis are used as bullet points / at paragraph starts. CoconutOctopus talk 11:38, 28 August 2025 (UTC)
• Oppose - changing my !vote after seeing GreenLipstickLesbian's rundown. Conditional oppose: Emojis are more indicative of being on a mobile phone than anything, and "no encyclopedic value" is more subjective than I would prefer for a speedy criterion. I would however support narrower criteria, like emoji bullet points. Gnomingstuff (talk) 16:18, 28 August 2025 (UTC)
• Oppose - in my experiments with CHAT GPT I have yet to see it use emojis. In my experience with humans, I have seen them use emojis. I also have had some browsers convert typos to emojis automatically. I don't think this is an obvious AI-tell that raises to the level of being added to the CSD. ~ ONUnicorn^{(Talk|Contribs)}problem solving 16:22, 28 August 2025 (UTC)
• Oppose. Firstly this is not a reliable indicator of unreviewed LLM output; secondly no evidence has been presented that this is needed: proponents should provide examples of pages that should be deleted, but cannot be deleted under the existing speedy deletion criteria (not just G15), but could be if something related to emojis were adopted and some evidence that these occur frequently. Thryduulf (talk) 22:16, 28 August 2025 (UTC)
• Oppose - literal years of debates over content with emoji show that people (particularly younger generations) have been using emoji in their natural writing for much longer than LLM content generators have been a thing, and many inexperienced editors write in the way that they naturally communicate rather than the fairly polished academic style of Wikipedia. Also per Thryduulf: this fails the "frequent" bullet of the CSD expansion criteria as no example use case has been identified at all. Ivanvector (^Talk/_Edits) 22:30, 28 August 2025 (UTC)
• Weak support, provided at the beginning of paragraphs or in place of bullet points. Emojis sprinkled willy-nilly through the text aren't an indicator of AI (though they will reflect other things about the editor), but the bullet point things is a hallmark. I would urge caution if this is the only criterion met though, and the text doesn't otherwise show signs of AI – the writer may just be an imbecile rather than an LLM. Cremastra (talk · contribs) 22:35, 28 August 2025 (UTC)
• Oppose Not a clear indication of AI use. Curbon7 (talk) 23:52, 28 August 2025 (UTC)
• Weak support. --📎 JackFromWisconsin (talk | contribs) 01:35, 29 August 2025 (UTC)
• I would oppose for failing the frequency criterion. I would also suggest freezing G15 for a few months with no further expansions, while it beds in. Stifle (talk) 08:06, 4 September 2025 (UTC)
• Oppose not obvious AI use. jolielover♥talk 17:18, 6 September 2025 (UTC)
• Oppose. Suspected AI contributions should be vetted carefully, not speedily. Urhixidur (talk) 17:43, 6 September 2025 (UTC)
• Oppose, not a sure tell for being AI, and not a good way to treat human volunteers. Emojis are easily removed through normal editing. ϢereSpielChequers 18:46, 6 September 2025 (UTC)
• Oppose as nonspecific. Also, if that's the only problem, it's easily WP:SURMOUNTABLE. WhatamIdoing (talk) 20:41, 6 September 2025 (UTC)
• Oppose - I don't think it's a clear enough tell of LLM misuse, doesn't seem frequent enough to be an especial problem that can't be dealt with through other means. Cheers, SunloungerFrog (talk) 06:20, 7 September 2025 (UTC)
• Support I agree with Cremastra. Polygnotus (talk) 13:51, 7 September 2025 (UTC)
• Oppose as mentioned above I don't think is something that should be used as criteria for a speedy deletion^Thanks,L3X1 ◊distænt write◊ 19:56, 7 September 2025 (UTC)
• Support - Although I do SUPPORT, a citation that gives evidence of LLM output being prone to use emojis would be helpful. I perused the linked WP:AISIGNS which supports this proposal. It also states in the second-to-last paragraph of the lead that "The speedy deletion policy criterion G15 (LLM-generated pages without human review) is limited to the most objective and least contestable indications that the page's content was generated by an LLM. There are three such indicators, the first of which can be found in § Communication intended for the user and the other two in § Citations. The other signs, though they may indeed indicate AI use, are not sufficient for speedy deletion." It is an WP:ADVICEPAGE only, not WP policy.--FeralOink (talk) 07:52, 9 September 2025 (UTC)
• Support Any human who has read a few Wikipedia articles knows they never use emoji (unless its a page about an emoji or in a quote) so if an AI makes a wikipedia article without being told its going on wikipedia, they wont follow WP:TONE microTato (talk) (contribs) — Preceding undated comment added 17:24, 9 September 2025 (UTC)

  I wonder if we all have the same idea about what an emoji is. How many of these bullets are emojis?

   • Item A

   • Item B

   ▪️ Item C

   ⚫️ Item D

   · Item E

   ∙ Item F

   ● Item G

   ❥ Item H

   ○ Item I

  None of these should be used to format a list in Wikipedia, but how many of these are actually emojis? WhatamIdoing (talk) 18:08, 9 September 2025 (UTC)

  On my device (Firefox on Windows 10), only C and D render as emojis; the rest are plain text. OutsideNormality (talk) 19:38, 9 September 2025 (UTC)

  At a glance on my device (Firefox on xubuntu linux) I wouldn't identify any of them as emoji but as normal text bullets. I could accept an argument that maybe D is an emoji, but what is the dividing line between an non-alphabetic (or equivalent in other scripts), non-punctuation character and an emoji? If we go by the unicode definitions, then I believe that none of the above are emoji. Indeed the only emoji used in this discussion are (in no particular order): 🐧💌🦋🐄🫘💕⭕✅🔹. The first 5 of those appear only in signatures, the 6th appears to be random text decoration in a human-written comment and the 7th is given as an example of legitimate use on another wiki. Thryduulf (talk) 21:54, 9 September 2025 (UTC)

  I just used text selection to determine if the examples are emoji or not. If the bullet turned white when I selected it, it was being rendered in plain text. If it stays black (or whatever colour it originally was) then it was being rendered as an emoji. That's also why I specified my operating system and browser, because emojis render differently on different devices. OutsideNormality (talk) 01:09, 17 September 2025 (UTC)

  None of them change color on my laptop (MacOS). I doubt that anyone here thinks that Item C (square dot emoji) is actually more obnoxious in an article than Item H (heart-shaped dingbat)?

  IMO since we don't have a shared understanding of which characters are emojis, and since we do have a l-o-n-g history of people typing things in word processing documents and then copy/pasting the results – something that's particularly noticeable when there's a bullet list that uses non-wikitext formatting – we really shouldn't be setting up a rule that says "I get to delete your contribution without discussion if you didn't use my preferred style of bullet". WhatamIdoing (talk) 22:30, 24 September 2025 (UTC)

• Support and also support adding "and headers" since in my experience ChatGPT uses them very often in headers. FaviFake (talk) 18:43, 13 September 2025 (UTC)
• Oppose Yes, LLMs are notorious for including emojis in their output, but emoji usage alone isn’t a definitive sign of AI writing; new editors sometimes use emojis in their writing. Without other signs of AI writing, it’s not objective enough for speedy deletion. Rosaece ♡ talk ♡ contribs 10:53, 17 September 2025 (UTC)

  I don't think new editors use emoji at the beginning of paragraphs or in place of bullet points. Note that this proposal isn't so broad as to include "emoji usage alone". FaviFake (talk) 14:34, 17 September 2025 (UTC)

  You may not think that, but people absolutely do. [3] [4][5][6][7][8][9][10][11][12]. All pre 2022-ChatGPT boom; a few might get speedied for other reasons such as copyright issues. GreenLipstickLesbian💌🦋 22:47, 17 September 2025 (UTC)

• Support as someone who always has and always will take a WP:TNT approach to this sort of thing. Remember that emojis alone aren't grounds for speedy deletion; this proposal will simply allow them as requisite evidence to call a spade a spade and kill a blatantly AI-generated page with fire under G15 (which isn't hard without any of these tells). Passengerpigeon (talk) 21:22, 17 September 2025 (UTC)
• Oppose. Faulty or bad use of emojis does not provide a definite conclusion that an article is written by LLM. Speaking from my own experience, chatGPT (and maybe other LLMs) are using less emojis these days.✠ SunDawn ✠ Contact me! 16:44, 21 September 2025 (UTC)
• Oppose. Would seem to make more sense to send the emoji title to its respective XfD board, then WP:SALT if necessary due to repeat creations. Steel1943 (talk) 18:56, 24 September 2025 (UTC)

• Oppose. Early this year I did a bit of online work for an AI-training company, so for a few months I was spending hours a day working with AI writeups, and I almost never saw emoji. I was instructed to watch out for emoji and flag them if they appeared, so I'm confident that output from the models with which I was working (I wasn't told which ones they were) would be highly unlikely to include emoji — therefore, to me, the presence of emoji indicates a human author. Nyttend (talk) 20:37, 26 September 2025 (UTC)

Discussion (Emojis)

• Just to clarify, I am completely open to any suggestions on changing the verbatim of the proposed clause. ~/Bunnypranav:<ping> 12:25, 27 August 2025 (UTC)

  What about something like "emojis for no encyclopedic reason". We shouldn't prohibit the use of it in articles like Emoji, where they serve a genuine purpose. Kovcszaln6 (talk) 12:57, 27 August 2025 (UTC)

  Done, thanks! ~/Bunnypranav:<ping> 13:01, 27 August 2025 (UTC)

  Do you have any diffs of emoji use in LLM content "in the wild"? -- LWG ^talk 13:38, 27 August 2025 (UTC)

  I unfortunately do not have any handy. The couple that I found were CSDd due to other G15 criterion, so we can't view them anyways. ~/Bunnypranav:<ping> 13:40, 27 August 2025 (UTC)

  Searching for these in draftspace is somewhat difficult, thanks to our tolerance of emoji in signatures and the ill-advised practice of putting {{AfC comment}} on the pages themselves instead of talk, and separately the use of emoji in reference titles (especially from the likes of Instagram); but for the sake of argument, an extreme low-effort search turned up Draft:Prowl transformer, Draft:Amy Sparkes, Draft:PATRICE MUBIAYI, and Draft:Dr. Arth White before I stopped looking. (Two of those are already clear speedies even without this proposed expansion, and I'd bet all four would be deleted if tagged; but please don't go and tag them just yet.) —Cryptic 16:54, 27 August 2025 (UTC)

  Not a page creation, but this ChatGPT experiment from Jimbo Wales shows emoji use. OutsideNormality (talk) 02:07, 28 August 2025 (UTC)

  Here's a clearly AI-written userpage with the typical emoji styling of such. Sarsenet•he/they•(talk) 12:14, 28 August 2025 (UTC)

• Second thought: even though emojis shouldn't be used in human-written encyclopediac content, it's certainly plausible that they might be, and I don't want G15 to slide towards "speedy delete anything that violates the MOS". If other signs of AI writing are present then emoji use is evidence that no reasonable human review happened but emojis aren't unambiguous signs of AI by themselves, are they? -- LWG ^talk 13:45, 27 August 2025 (UTC)

  Hmm. I do not think emojis are plausible in human text, unless at least some of it is copied from AI. The criterion specifies [...] removed by any reasonable human review anyways, so this shouldn't go that extreme of deletion due to violation of MOS. ~/Bunnypranav:<ping> 14:19, 27 August 2025 (UTC)

  I do not think emojis are plausible in human text In encyclopedic content, probably not. But as general "human text", I've been increasingly using emojis as tone indicators in online discussions. See also Emoji#Linguistic function of emoji. Anomie⚔ 14:38, 27 August 2025 (UTC)

  What about narrowing it down only to emojis in front of section headers or as bullet points (✅ and 🔹)? They are the main places where AI uses them, and based on my quick read of your comment, humans may put it in the content/end of sentences (correct me if I'm wrong). ~/Bunnypranav:<ping> 14:43, 27 August 2025 (UTC)

  I agree with this limitation to adding emojis as a G15 criterion and have added my conditional support above accordingly. ViridianPenguin🐧 (💬) 11:36, 28 August 2025 (UTC)

  I mean, there's a reason we've had an edit filter since 2015[13] to prevent new editors from adding emojis to articles. Anecdotally, I've seen people use emojis in ref names, pre-chat GPT. I can't dig out the example I'm thinking of but I remember seeing somebody doing that in pre-2020 page revisions. (And then being very sad when I discovered that somebody had removed them from the modern version). I get the feeling behind this proposal, but it needs a lot more workshopping to be viable. Here's a few examples of emojis/unicode stuff being used in articles, anywhere from constructively to "plausibly good-faith".

  • Probably a decent number of the quarter million hits on Filter 680, actually.[14]. Many of them probably aren't going to be worth saving, but very few are going to be unambiguously LLM generated, and more than a few will be false positives. But yeah, this does prove that humans plausibly add emojis to articles.
    • [15] and [16] for a couple seemingly good-faith "no, we don't want you to do that edit but it's not exactly wrong either" style hits.
    • [17] Edit filter person talking about how you have to exempt Japanese-related articles from emoji/unicode filters like this.
  • Check and x emojis in like tables and stuff to signify yes/no. For example, List of works produced by Hanna-Barbera Productions, Freeze brand, and 1922 Swedish prohibition referendum.([18][19][20] diffs where emojis where added, for the curious).
  • [21] Globe emoji in a table
  • Skull and cross bones emoji is in List of stock characters, but it's hidden behind a piped link. (and probably should be removed)
  • Since 2014, a skull a cross emoji has represented that somebody died in Anti-cession movement of Sarawak.
  • Planets in astrology uses a lot of decorative emojis.[22] I think they might be astrology symbols, but I really don't know. To an admin not versed in astrology, these have no value, and a human certainly should not have left them in past review. To one versed in astrology, I'm assuming these are plausible, but they shouldn't have been added anyway.
  • We sometimes have issues with people trying to add these to short descriptions (probably the low character count makes them think it's social media, idk)[23][24].
  • Sports editors gonna sports editor[25] + further conversation about how newbies try to mimic images/symbols by adding emojis to articles[26]
  • [27] 2018 warning from an admin to a new editor about using emojis in articles. It looks like the emoji was in a ref name to, so for anybody saying that "obviously, an admin wouldn't speedy a draft because they copied the source title... there you go! (Waves a small trout at @Acroterion for historic mistakes, and a very genuine thank you for the data point)
  • The male/female emojis get used a fair bit[28], both usefully and not. For example, Hotung family, anything transcluding Template:Pharaohs, this random dab page Ailing (Chinese name).
  • [29] This newbie drafting a list to potentially add to an article on the talk page, using emojis in a decorative manner (all wordings we've come up with would kill this list, despite it being the most obviously good-faith one so far)

  • Similarly, [30] is a 2020 edit using the blue dot as a bullet-point esque marking. It's still in the article and very much predates chatGPT.

  In terms of what other wikis use (and therefore, what good-faith newbies will often add to our articles either by translation or by making the mistake of assuming what's okay there is okay here):

  • [31] zhWiki uses emojis in some templates
  • On jaWiki, ⭕ sometimes gets used for article like this

  Also, I can't find any diffs, but here are some other ways I can see people using emojis in a non-LLM generated fashion:

  • Some people add their edit summaries to the actual article, especially in draft space. Some people might use emojis in those comments. If somebody ran their draft through something like Grammarly, it's probably going to have LLM vibes and the fact that somebody added emojis should not make it speediable.
  • Stuff like the skull and cross emoji or hazard emojis ☢ are used in botany and chemistry to denote that plants or chemicals are scary, so it's plausible a newbie might try to add those symbols to article text in what would appear to be a decorative way.
  • flaggggggsssss.

  I'd imagine that AfC editors tend to mostly see emojis in LLM-generated spammy draft, but I'm not convinced that if anybody had instead of making these additions to an existing article, started out a new article and included emojis in such a manner, then their work should be insta-vanquished. "encyclopedic value" is nebulous; I'd argue that very few of the examples I listed actually have any value, and I'd happily remove most of the emojis if I wasn't using them as examples. But no, I don't agree with the premise that emojis cannot be plausibly inserted by a human. GreenLipstickLesbian💌🦋 21:47, 27 August 2025 (UTC)

  Actually, as far as those "this emoji seems decorative/valueless to some editors but isn't really", this edit by a newbie was stopped in the edit filter because they added an unicode emoji heart thing[32]. They tried to get somebody to add the emoji for them, but were turned down [33] and later templated by an admin for MOS:DECOR issues relating to the edit.[34] What neither of the more experienced editors realized, however, despite the newbie explaining themselves and the fact there were already 25 other hearts on the page and that fact it's explained in the key, is that for some esoteric reason Wikipedia editors use the heart emoji to denote, I believe, mares.[35]. So as far as "encyclopedic value" of emojis, again, not as objective as it might appear on first glance. (Also hi @DreamRimmer and @Peaceray! I'm talking about some random edits you two made a few months ago, so sorry but also thank you providing me with an example!)GreenLipstickLesbian💌🦋 00:09, 28 August 2025 (UTC)

  Apologies, I will try to be more careful next time. – DreamRimmer ■ 13:27, 28 August 2025 (UTC)

  Oh gosh, no worries at all, and mostly just a giant thank yo for manning the edit filter thing that reduces the number of headaches the rest of us editors have to deal with!💕 And yeah, I was born to a former horse-girl and a wanna-be cowboy, and it took even me a minute of thinking and reading to figure out the heart emoji thing. GreenLipstickLesbian💌🦋 13:32, 28 August 2025 (UTC)

  Haha, you are definitely a genius, and I have often noticed your great skills in handling copyright-related matters. I also want to thank you for all the wonderful work you do. – DreamRimmer ■ 13:40, 28 August 2025 (UTC)

  Thank you GreenLipstickLesbian for your detailed comment with diffs and examples! Based on your, and some others' suggestions, I have narrowed the proposal down specifically to bullets and start of paras. Wondering if you have a new take on this, or any further amendments for more clarity and less false-postives. ~/Bunnypranav:<ping> 13:56, 28 August 2025 (UTC)

  Sorry for the delay, wanted to think about my response a bit more! I definately prefer the much more narrow scope and I'm a lot less opposed to it, however, I still have concerns. For example, I've just quickly found this userpage, created in 2020, that uses bullet point emojis instead of actual bullet points. The userpage or it's history is going to vanish at some point, but that's because it was a verbatim copy of this WHO Facebook post[36]. So obviously not LLM-generated, and this is an example of not one, but likely a small team of humans looking at text with emoji bullet points and going "yeah, seems good" followed by a new editor doing the exact same thing. I admint this is a slightly weird case, as in this page could be validly CSD-ed, but I don't think anybody wants this CSD-ed under G15.

  I'm also a little worried that some new editors (especially in poorly-thought through editathons or classes) might use something like ChatGPT to generate an outline, then fill it in with (possibly quite poorly written) human text. I can't remember the exact pages, but I'm pretty sure I saw a few drafts like that yesterday which were part LLM-placeholder with emoji titles, but the actual writing could have only been written by a human. (Part of the reason I didn't save the titles is that both articles sucked in a uniquely organic way, so I couldn't use them as examples as I don't like public shaming an unsuspecting human's writing if I can help it) And, I mean, as far as using AIs to edit Wikipedia.... I mean, I wouldn't recommend doing that for obvious reasons but it's something an otherwise competent human might plausibly do, it's not actually the worse way to use an LMM to create articles, and I think speedying those could be bitey and create a headache at DRV. I mean, a decent portion will suck and be horrible, but will they be horrible enough that I don't mind them acidentally getting cleared out under G15? I don't know. @Bunnypranav GreenLipstickLesbian💌🦋 21:51, 28 August 2025 (UTC)

• As with my comment in #RfC: Including Markdown in G15 above, it's not emojis per se (or markdown, or dead links) that should make a page speedy-deleteable; it's subjectively LLM-generated content which is also confirmed by one of these blatant signs. —Cryptic 15:11, 27 August 2025 (UTC)

  I agree with this, although that's not currently clear from how G15 is written. -- LWG ^talk 16:00, 27 August 2025 (UTC)

• Just noting that blatantly unencyclopedic emojis can come from word prediction/autocorrect systems, especially if the emoji comes directly after a corresponding word. E.g. Draft:CMantham and Draft:David E. Reed appear to have emojis come from word prediction, not an LLM. Helpful Raccoon (talk) 22:15, 27 August 2025 (UTC)
• Since the markdown RfC is on WP:CENT, should this one also be listed there? OutsideNormality (talk) 04:16, 28 August 2025 (UTC)
• Notified: WT:AFC, WT:NPP, WP:CENT. ~/Bunnypranav:<ping> 05:44, 28 August 2025 (UTC)

  Would notifying WP:AIC be considered canvassing? Children Will Listen (🐄 talk, 🫘 contribs) 13:33, 28 August 2025 (UTC)

  Hope not. GreenLipstickLesbian💌🦋 13:38, 28 August 2025 (UTC)

• Per various suggestions above, I have amended the proposal to only be applicable to main and draft page, that too only in start of paragraphs of in place of bullet points. Open to rephrasing in any way for clarity. ~/Bunnypranav:<ping> 12:09, 28 August 2025 (UTC)

  @Gnomingstuff Would you be willing to suggest an alternative for "no encyclopediac value"? Also, as of now, the proposal limits to bullet points and para beginnings, which I assume has no encyclopediac value by default. ~/Bunnypranav:<ping> 16:49, 28 August 2025 (UTC)

  I think the narrowed criteria are fine. Gnomingstuff (talk) 17:14, 28 August 2025 (UTC)

• I've definitely seen this. Shallow web was an example that was deleted on other grounds. As I recall, the original version had emoji for bullet points along with other AI tells. The emoji were cleaned up but the other tells remained. Stepwise Continuous Dysfunction (talk) 11:36, 31 August 2025 (UTC)

  deleted on other grounds [and also had] other AI tells so its not an example of a criterion like this being needed. To demonstrate that there needs to be many examples of pages which cannot currently be deleted under existing criteria (including but not limited to G15). Thryduulf (talk) 12:18, 31 August 2025 (UTC)

  If we can just send everything to AFD, then no speedy deletion criteria are needed. Sheesh. And no, this cannot currently be deleted under existing criteria. —Cryptic 12:31, 31 August 2025 (UTC)

  Speedy deletion is the exception not the rule. Only if there are sufficiently many instances of the same sorts of pages getting nominated and always being deleted is speedy deletion appropriate. You also claim that it both had multiple AI tells and no other AI tells, it cannot be both. Thryduulf (talk) 14:42, 31 August 2025 (UTC)

  G11 has no specific criteria and it still works pretty much all the time. Why can't we just interpret G15 like G11? Children Will Listen (🐄 talk, 🫘 contribs) 14:46, 31 August 2025 (UTC)

  Advertising is more reliably identifiable than AI slop. —Cryptic 14:55, 31 August 2025 (UTC)

  G11 is based on a page's state (the state of being literal spam), while G15 is based on the page creation process (i.e. that it has been generated and slopped onto Wikipedia with no human review), which is harder to determine objectively: at best, we can make specific criteria (the G15 criteria) which indicate that the page in question likely had no human review, but we can't time-travel to when the creator created the page. We can't just redefine G15 to be all potentially AI-generated content either, as not all AI-generated content is bad enough that we need to skip a deletion discussion and speedily delete it (not to mention false positives). OutsideNormality (talk) 17:03, 31 August 2025 (UTC)

  Speaking of G11, I sometimes wonder what we would learn if we compared FAs for major businesses against Wikipedia:Identifying blatant advertising#Typical signs of blatant advertising. Until recently, it said that you could identify blatant, CSD-worthy advertising because it mentioned when a company was founded. WhatamIdoing (talk) 20:49, 6 September 2025 (UTC)

  It provides an external link to the company's online store, the product's homepage, marketing info or specifications page, product purchasing order page .... yes, that's killing most of our articles on businesses.[37][38][39] And I have yet another thing to add to my list of "fine if an older editor adds it, blockable offense if it's in somebody's first ten edits" [40] (the block was discussed at length on AN so I don't feel too bad about talking about it here) GreenLipstickLesbian💌🦋 22:11, 6 September 2025 (UTC)

  I've long said that Wikipedia's rules apply less to editors like me than to some other editors. This makes some sense: For example, I am the all-time top editor of Wikipedia:External links, its talk page, and the associated noticeboard,[41][42][43] so why shouldn't you trust me on a questionable URL? The plain fact is that there probably isn't anyone else on the planet who knows these rules better than I do, and fewer than I'd like who equal my knowledge. But there is also something slimy about a system, created by me and people like me, that says "we" are permitted to add article content that a new editor would be reverted or even punished for. WhatamIdoing (talk) 17:54, 9 September 2025 (UTC)

  The "tells" in the article text did not fall under G15. The AfD nomination argued that it was AI-generated, not because of any of the G15 criteria, but because there were websites and programs that seemed to have been completely made up. In other words, the nomination invoked "other, more subjective signs of LLM writing that [...] should not, on their own, serve as the sole basis for applying" G15. Stepwise Continuous Dysfunction (talk) 22:31, 31 August 2025 (UTC)

  "subtle, subjective signs" are not suitable for speedy deletion because speedy deletion criteria must be objective and obvious. Thryduulf (talk) 22:36, 31 August 2025 (UTC)

  There was nothing either subtle or subjective about the use of emoji on that page. —Cryptic 23:10, 31 August 2025 (UTC)

  But, as repeatedly noted in this thread, that is not a reliable indicator of unreviewed LLM output. Thryduulf (talk) 23:14, 31 August 2025 (UTC)

  I wasn't saying that it should have been speedied because of the problems that the nominator pointed out. I'm saying that it's an example of a page that was identified as AI-generated, which was also full of emojis in exactly the way that has been proposed as an indicator of AI-generated text. Stepwise Continuous Dysfunction (talk) 20:10, 1 September 2025 (UTC)

• If we cast too wide a net, we will catch articles that just have odd formatting. I recently saw an edit summary from an excellent and experienced editor, where the edit was largely a positive change, but it noted "I'm going to assume that this list of four came from AI." However, that list formatting was present in the Spanish Wikipedia article back in 2018, and reflected the source (Nores et al., 2015) cited there. Rjjⁱⁱⁱ (talk) 01:34, 4 September 2025 (UTC)
• Emojis added before ChatGPT became a thing should be excluded. Zero^talk 01:52, 7 September 2025 (UTC)
• I think the G15 should instead acknowledge that bad formatting is a possible sign of LLM use, but it should not be the only indication, depending on context and the editor's experience with wikitext. I am also thinking about how G15 could note the existence of Wikipedia:Signs of AI writing while clarifying that it is an advisory guideline. --Minoa (talk) 09:38, 9 September 2025 (UTC)
• Oppose: This criterion only applies to absolutely, entirely unambiguous cases of LLM use. There are plausible reasons why a good-faith human editor would use emoji. JJP_Master (she/they) 12:21, 11 September 2025 (UTC)

  Did you mean to put this in the section #Survey (Emojis) above? OutsideNormality (talk) 01:00, 12 September 2025 (UTC)

Side proposal

G15 is new, it seems to be working reasonably well thus far, but it's still bedding in and there's been a plethora of proposals, many on CENT, to add or tweak the criterion. I propose a 2-month moratorium on proposals to make further tweaks to the criterion while this bedding-in takes place, so that it isn't constantly changing during that time. After that point we will have a clearer view on what needs to be added or changed. Stifle (talk) 08:10, 4 September 2025 (UTC)

This is not a bad idea. It takes time to figure out whether/how a rule is working, and if you change it every few weeks, you'll never figure it out. WhatamIdoing (talk) 21:15, 6 September 2025 (UTC)

I think that this is a good idea. G15 now allows editors to deal with egregious cases of LLM misuse quickly. Rather than agonising over less obvious cases or criteria, let's let it settle down. Nothing to stop people from noting down personally cases that might allow the G15 umbrella to be expanded, for future discussion. Cheers, SunloungerFrog (talk) 06:16, 7 September 2025 (UTC)

New speedy deletion criteria

The following discussion is closed. Please do not modify it. Subsequent comments should be made on the appropriate discussion page. No further edits should be made to this discussion.

---

I propose that Wikipedia adds another speedy deletion criteria called "R5. Cross-namespace redirect from namespaces other than the mainspace".

It will apply to redirects such as:

• User to template namespace redirects
• Talk to Wikipedia namespace redirects
• Portal to user namespace redirects

though not stuff like:

• Mainspace to user namespace redirects
• Mainspace to draft namespace redirects

which violates R2 instead. BodhiHarp (talk) 18:14, 27 August 2025 (UTC)

Oppose. Either not frequent enough to warrant speedy deletion rather than RfD or no reason why they should be deleted in the first place. * Pppery * _{it has begun...} 18:25, 27 August 2025 (UTC)

Cross-namespace redirects are often kept at RfD. Of the three examples you give, the first will usually be harmless (although on a top-level userpage it may need to be softened per WP:CATTEMCATREDIR); the second would almost certainly be an error that can be fixed with G6 deletion or simply replacing the redirect with {{talk header}}; and the third would probably also be a G6able error, and if not is something that RfD could handle. -- Tamzin^{[cetacean needed]} (they|xe|🤷) 18:26, 27 August 2025 (UTC)

I don't believe such problematic cross-namespace redirects occur often enough to justify a new criterion. Most of these that are created in error are already eligible for deletion under WP:G6 or (in the case of talk pages) WP:G8, and the few exceptional cases won't overwhelm WP:RFD. ^Complex/_Rational 18:30, 27 August 2025 (UTC)

@BodhiHarp: Please see the box shown to you when posting here, beginning "Before proposing or commenting on new or expanded criteria", which is a summary of the larger box at WP:NEWCSD. While you appear to have satisfied (1) Objective, I don't see that any of the other three (Uncontestable, Frequent and Nonredundant) have been met.

A general principle, that should be followed by anybody seeking to create a new rule or amend an existing rule, is: can you justify that new rule or amendment in a non-hypothetical form? In other words:

1. demonstrate that a problem exists
2. show that existing processes are insufficient to overcome the problem
3. propose a method that will either solve the problem, or prevent it from re-occurring
4. invite comments and be prepared to defend your proposal
5. if sufficient parties are in agreement, implement the proposal

You've come up with (3) and (4), but not (1) or (2). To satisfy (1), please provide several examples of redirects (they may be any or all of: (i) existing and not at WP:RFD; (ii) presently at RfD; or (iii) previously deleted at RfD) to which this proposal would apply. The more examples that you can provide, the easier it will be for us to observe that a problem exists that needs to be fixed. If it's happening on a more-than-daily basis, we probably do need to take action; but if it's only once a year, it should be quite sufficient to follow the existing RfD process. --Redrose64 🌹 (talk) 07:27, 28 August 2025 (UTC)

Oppose as I can think of a number of exceptions that are the result of moves. And a dozen examples of problems where this criterion would be useful has not been provided.Graeme Bartlett (talk) 07:40, 28 August 2025 (UTC)

No per tamzin. 📎 JackFromWisconsin (talk | contribs) 01:39, 29 August 2025 (UTC)

Oppose per Tamzin and Redrose64. Thryduulf (talk) 17:01, 30 August 2025 (UTC)

Oppose even though I proposed it: I realize that if this was added, redirects from drafts to the mainspace would be deleted. BodhiHarp 18:56, 10 September 2025 (UTC)

The discussion above is closed. Please do not modify it. Subsequent comments should be made on the appropriate discussion page. No further edits should be made to this discussion.

Merge F9 and G12

The following discussion is closed. Please do not modify it. Subsequent comments should be made on the appropriate discussion page. No further edits should be made to this discussion.

---

I think that "F9. Unambiguous copyright infringement" should be merged with "G12. Unambiguous copyright infringement". BodhiHarp (talk) 16:13, 30 August 2025 (UTC)

Oppose. Copyright violating text (G12) and copyright violating files (F9) have very different considerations. For example, fair use is very relevant to files but not really to text; close paraphrasing and non-violating earlier versions are obviously irrelevant to files but something that needs to be considered every time for text. There is actually relatively little overlap in the detail of the criteria once you read beyond the title, and most of what's there for each would need to be retained if merged leading to a more complicated criterion but no real benefit. If any changes are needed it's to the titles - G12 → Unambiguous copyright infringing text, F9 → Unambiguous copyright infringing files. There would be both a small but not non-existent benefit and a small but not non-existent amount of disruption from such a change, I'm not sure which outweighs the other. Thryduulf (talk) 16:59, 30 August 2025 (UTC)

See Wikipedia:Perennial proposals § Merge speedy deletion criterion F9 into G12. They were previously a single criterion, and it was very unwieldy. HouseBlaster (talk • he/they) 19:04, 30 August 2025 (UTC)

Oppose. Different skills, different watchers. SmokeyJoe (talk) 23:21, 30 August 2025 (UTC)

Oppose per above. Headbomb {t · c · p · b} 23:36, 30 August 2025 (UTC)

Oppose per the link to prior discussions provided by HouseBlaster. -- Whpq (talk) 00:07, 31 August 2025 (UTC)

Maybe a footnote in F9 mentioning that, while it's superficially covered by G12, it's deliberately kept separate as a matter of practicality? I wouldn't object to Thryduulf's proposed title changes either. —Cryptic 01:13, 31 August 2025 (UTC)

The discussion above is closed. Please do not modify it. Subsequent comments should be made on the appropriate discussion page. No further edits should be made to this discussion.

Rename F9 and G12 headings?

Should we clarify the headings of F9 and G12 to emphasise the difference between them? Something like

• G12 → "Unambiguous copyright infringing text" or "Unambiguous copyright violations (text)"
• F9 → Unambiguous copyright infringing files or "Unambiguous copyright violations (files)."

This would hopefully add clarity and reduce the number of proposals to merge them. There would be both a small but not non-existent benefit and a small but not non-existent amount of disruption from such a change, I'm not sure which outweighs the other but given Cryptic said they wouldn't be opposed, I figured it's worth seeing what others think of it independent of any other suggestions (there would be no changes to the text of either criterion, or anything else other than the heading). Thryduulf (talk) 15:06, 1 September 2025 (UTC)

I'm not sure this would do it. Those proposing the merge of the two are aware that one covers text, and the other covers files. What they are likely not aware of is the reason they are separate, and the past discussion. Cryptic's suggestion of a footnote would work better. The footnote could refer to the perennial proposal link. -- Whpq (talk) 12:32, 2 September 2025 (UTC)

I think that renaming the headings could help people who are looking at the Table of Contents or links to specific sections, especially if they don't understand our codes (G for General, F for Files). WhatamIdoing (talk) 21:18, 6 September 2025 (UTC)

Formality and counterproductivity of G5

The G5 criterion strikes me as the most counterproductive and formal criteria of SD. If an article is perfectly healthy (verifiable, reliable, netural), who cares whether the major editor is banned, blocked etc.? By deleting it we create extra work for other users who should recreate the article from scratch. This seems to me to be against Wikipedia's Five pillars.

Plus no rationale is provided in the G5 description of why speedy deletion must be enforced.

It seems that in essence G5 is a retributive, namely a way to get back at transgressing users, which is nothing but harmful. Xpander (talk) 09:43, 2 September 2025 (UTC)

It's not a must. The administrator can decline the tag and claim responsibility for the content. That causes the page to stop being eligible. Any user who would delete (administrator) or tag can put in a dummy edit or leave a talk page message to the same effect; if the page gets deleted nevertheless, they can request undeletion because G5 had not applied, and the deletion was an error. —Alalch E. 13:58, 2 September 2025 (UTC)

I wrote an essay on this topic: Wikipedia:G5 is not a firm rule. Mz7 (talk) 23:14, 5 September 2025 (UTC)

Very good essay. SmokeyJoe (talk) 00:24, 6 September 2025 (UTC)

Thanks for the essay, why is there no mention of it in Speedy Deletion? @Mz7 Xpander (talk) 06:06, 7 September 2025 (UTC)

Policies might sometimes link to common essays but this is avoided so as not to cause the essay to be cited as the only correct interpretation of a policy and be applied effectively instead of the policy; someone will then say "it's just an essay", which will be replied to with "yes, but it's linked from the policy page itself, so that essay is what the policy actually means, it's the 'policy-endorsed' essay"—this is a recipe for intractable conflict, and policies exist precisely to reduce conflict. For an essay to be linked in the way you suggest, from a policy, it should be a tm:supplement, a.k.a wp:explanatory essay. Documentation says Use this template carefully, only when there is a well-established consensus at the relevant policy or guideline page to use this template on an essay that links from the relevant policy or guideline. This might be that discussion. And indeed, I support retagging that essay as a supplement, because it is correct on policy, it does not contradict the very text of the policy, it reflects the preferred understanding of the policy, it describes best practices, clarifies principles, and resolves conflicts around G5, and is neutrally written. —Alalch E. 18:46, 7 September 2025 (UTC)

It is not correct on policy, it conflicts with Wikipedia:Banning policy (the section suggesting it doesn't redefines banning as "will likely still be blocked", whereas the actual policy details how the edits themselves are presumed harmful). CMD (talk) 16:58, 8 September 2025 (UTC)

I don't quite understand your argument that my essay conflicts with policy. The central point of my essay is that G5 does not require administrators to delete qualifying pages unconditionally. The most relevant part of the policy is WP:Banning policy#Edits by and on behalf of banned and blocked editors, which states: This does not mean that edits must be reverted just because they were made in violation of a block or ban (changes that are obviously helpful, such as fixing typos or undoing vandalism, can be allowed to stand), but the presumption in ambiguous cases should be to revert. This clearly supports my essay. There is a section WP:Banning policy#Bans apply to all editing, good or bad—nothing in that section says we should unconditionally revert edits that are obviously good, only that we should still block editors for violating bans even if the ban-evading edits were good. Mz7 (talk) 03:48, 9 September 2025 (UTC)

G5 does not require admins to delete qualifying pages unconditionally because there is no situation that requires administrators to delete qualifying pages unconditionally. At any rate, the conflicting part is the interpretation of what banning means. A ban does not mean someone is likely to be blocked, it means they are specifically prohibited from any form of editing. Administrators are obviously not "required" to implement this "unconditionally", but again that applies to all policies. CMD (talk) 07:11, 9 September 2025 (UTC)

Which part of the G5 text implies that it's not a mandate or not required? There's no unless. @Chipmunkdavis Xpander (talk) 13:11, 9 September 2025 (UTC)

Admins are not mandated or compelled to take any particular action. They are, to an extent, mandated to account for admin actions they take, but there is no mechanism (not community will for such a mechanism) that can mandate an admin to take an action. CMD (talk) 15:19, 9 September 2025 (UTC)

So even if an admin makes a mistake e.g. deletes your page by mistake, then they aren't obliged to restore it? i.e. to undo their mistake? Do you mind pointing me to a reference which describes the admins' non-obligation part? Xpander (talk) 15:53, 9 September 2025 (UTC)

The second paragraph at Wikipedia:Administrators includes the sentence They are never required to use their tools, and must never use them to gain an advantage in a dispute in which they were involved. This includes never being required to undo their own mistakes, because it is sometimes a matter of opinion whether something was a mistake. In the case of genuine mistakes every administrator should be happy to undo it themselves or ask another admin to undo it for them. If they disagree with or don't respond to a request then discuss the matter. Thryduulf (talk) 16:18, 9 September 2025 (UTC)

@Xpander1 the very first sentence of the speedy deletion policy begins: Speedy deletion is a process that provides administrators with broad consensus to bypass deletion discussion, at their discretion (emphasis mine). Nothing in the text of any criterion contradicts that. Additionally, the wording of the policy makes it clear that speedy deletion may be contested by anyone (in some cases excluding the page creator) and must only be used in "the most obvious cases". Thryduulf (talk) 15:40, 9 September 2025 (UTC)

Thanks for pointing that out. I now figured out that the first line is indeed ambiguous to me. What does "admins with broad consensus" mean?

Does it mean admins that have been elected by broad consensus? or Does it mean when there is a broad consensus to delete a page (where? in the admin's mind?), in that case shouldn't that be up for a vote? @Thryduulf Xpander (talk) 15:51, 9 September 2025 (UTC)

I only quoted part of the sentence. What it means is that there is broad community consensus that administrators may delete pages that unambiguously meet one or more of the listed criteria without first seeking consensus about that page individually. Thryduulf (talk) 16:00, 9 September 2025 (UTC)

(to Chipmunkdavis) Hmm, I'm still not sure I understand where you think my essay conflicts with policy. It seems like you and I fundamentally agree that G5 is not the unconditional mandate that some editors think it is (which may seem obvious in light of WP:IAR and WP:5P5, but it is nevertheless a common misconception—see the very post that spawned this thread). I am aware of what the difference between a "ban" and "block" is on Wikipedia, and I feel like my essay respects that difference: a block is the technical means by which a ban can be enforced (not all bans need to be enforced with a block; e.g. a topic ban would only result in a block if the user violated the topic ban, and WP:BMB refers to the idea that we would still block the user for violating the topic ban even if the violating edit was "good"). Mz7 (talk) 08:11, 10 September 2025 (UTC)

It is very short-sighted to view G5 as retributive. Many sockfarms are banned for good reasons, including harassment. Telling them that they're not actually banned, and that instead they can keep editing through new accounts, is deeply harmful to the community and undermines the very safeguarding purpose bans can have. CMD (talk) 00:17, 6 September 2025 (UTC)

Let's not throw the baby with the bathwater. Shall we? What I mentioned was about indiscriminate deletion of perfectly healthy or good articles, not about harassment. Xpander (talk) 06:09, 7 September 2025 (UTC)

"perfectly healthy or good articles" and there's the issue. No, if somebody was blocked as a VOA when they were a kid and comes back as a proper editor, then G5-ing their creations would be (imo) stupid. However, if somebody was blocked for serious content issues, especially concerning source falsification, plagiarism, UPE, POV-pushing, copyright issues, then their articles look okay at first glance. They might even get several admins nd very experienced users defending them because everything looks okay, isn't G5 just counterproductive, why are we depriving our readers of useful content, etcetera, etcetera, only to have the articles fall apart the moment you actually look at them. I don't know about you, but I value encyclopedia content and our readers more than the creations of somebody who we know is unwilling or unable to abide by core content policies, especially to the point where they were blocked. I value the people, especially people living in areas with minimal freedom of the press and poor regulations of undisclosed advertising who use Wikipedia to learn about the world above preserving suspect content written by somebody who was deliberately evading a block designing to stop them editing to make content designed to deceive others. (Which, in some extremes, involves pushing very subtle disinformation about ethnic groups or regions in order to support human rights abuses/genocide against them because people are just monsters sometimes) I mean, if you see an article that was g5 tagged, then by all means, go ahead and clean it up. You have access to the edit button, just like everybody else on this site. GreenLipstickLesbian💌🦋 19:43, 7 September 2025 (UTC)

It's the opposite of indiscriminate, it's a tightly defined CSD. As for what we shall do, perhaps not redefine blocks and bans to mean 'actually still able to edit and we are all okay with that'. CMD (talk) 16:52, 8 September 2025 (UTC)

This should be in the FAQ somewhere. Without G5, we would not have a banning policy. Banned means banned. —Kusma (talk) 16:59, 8 September 2025 (UTC)

• Xpander1 A block/ban is meaningless if a user can just evade it and create good content that remains. If they want to create good content, they should work to get unblocked and tell the good content they want to create. As noted, it's not required that deletion occur(very little is a "mandate" here); if you want to take on the task of following G5 tags around and evaulating them to see if the content is worth keeping, you are free to do so. It's not only admins that can remove G5 tags. 331dot (talk) 13:24, 9 September 2025 (UTC)

  One danger there is that LTAs may see that and actually be encouraged to evade their block/ban, knowing that you will rescue whatever they wrote. Keep that in mind. 331dot (talk) 13:28, 9 September 2025 (UTC)

Request for clarification on R3 for redirects not in the Latin alphabet

R3 includes the wording However, redirects from common misspellings or misnomers are generally useful, as are some redirects in other languages. and {{db-r3}} says [...], which is not in another language pertinent to the topic. This results in editors tagging redirects that (they think) fail WP:RLOTE for deletion per R3 (example) even when they are neither misspellings nor misnomers. The way I interpret R3's mentioning of foreign redirects is that something that looks like a misspelling of the English name, may in fact be the spelling of the name in another language. So tagging, for example, AbyneAbyne for deletion per R3 would be wrong because it's the Scots name of a village in Scotland and not a misspelling of Aboyne. On the other hand, tagging esagono -> hexagon (wikt:esagono) for deletion per R3 would be fine (?) because there is no special connection between the Italian language/culture and hexagons (not one I'm aware of at least) and therefore this redirect can be thought of as a very bad misspelling of the English "hexagon".

However, if a redirect is not written in the Latin alphabet then no such argument could be made. So I think εξάγωνο -> hexagon (wikt:εξάγωνο) would not qualify under R3. In such cases, the only way to delete per R3 would be for the term to be misspelled or be a misnomer in the foreign language. Is this correct? If not, then why does R3 apply here? Warudo (talk) 18:37, 2 September 2025 (UTC)

@Warudo, here's a simpler explanation: Redirects from (unrelated) foreign languages should normally be handled under WP:RFD#DELETE #8. The main exception is if you didn't realize it was a foreign language word and genuinely (though mistakenly) believed it was a English-language typo (and it was created in the last few weeks, because R3 only applies to "recently created" redirects). Your example of esagono doesn't fall under R3 because it doesn't look like (and really isn't) a misspelling. A redirect from that word should go through RFD.

See also Wikipedia:Redirects in languages other than English.

Remember that the goal of CSD is never to maximize the percentage of deletions that happen via CSD. Using a non-CSD process is a good thing. WhatamIdoing (talk) 21:43, 6 September 2025 (UTC)

@WhatamIdoing Ok then, so R3 cannot be used to tag bad WP:RLOTEs at all unless they are misspellings or misnomers in the language they are written. That makes sense to me. On the other hand, it still doesn't change the fact that people do it anyway and they do it without even checking whether the foreign language is relevant to the topic first. {{db-r3}} as written implies that it can delete foreign redirects if they are in a language that is not pertinent to the topic. I think it should be rewritten to something like:

[...] as a recently created redirect from an implausible typo or misnomer. Note that what may look like a misspelling to you might instead be the correct spelling in another language.

This makes it clear that R3 cannot be used to delete foreign redirects just because they are foreign, even if they are in a language that is irrelevant to the topic. Warudo (talk) 09:22, 7 September 2025 (UTC)

I'm not sure that "people do it anyway". I looked at the last 10,000 entries in the deletion log. There were 41 total deletions naming R3 (including those that were R3 plus some other reason).

I saw a small number that I thought were borderline as R3 (e.g., @Significa liberdade deleted The Roman Catholic Archdiocese of Toronto, when we have an article at Archdiocese of Toronto and a redirect from Roman Catholic Archdiocese of Toronto – is the presence of "The" an "implausible typo", or is it just a violation of Wikipedia:Naming conventions (definite or indefinite article at beginning of name) and therefore should have been handled through RFD or ignored?)

I saw none that involved a simple redirect from a foreign language. WhatamIdoing (talk) 18:27, 7 September 2025 (UTC)

You didn't see them in the deletion log because they get contested. I've linked an example of this happening in my initial comment. I have two more examples of this. Since speedy deletion is one of the most WP:BITEY processes we have, I consider incorrect tagging to be a problem as well, not just incorrect deletion. Warudo (talk) 18:41, 7 September 2025 (UTC)

That means that I don't see them in the deletion log because people don't do it anyway. What you've shown is that one (1) editor incorrectly believed that R3 could be used for a redirect in Japanese, to an article about the Japanese song of that same name. One person making a mistake and then getting reverted isn't a good reason to change the policy.

I think this was an instance of Wikipedia:Nobody reads the directions. The directions already say that some redirects from non-English names are good. Someone who read the directions probably wouldn't have tagged those pages for speedy deletion. And even when they do, other editors contest it, or the admins will refuse to delete them. WhatamIdoing (talk) 19:05, 7 September 2025 (UTC)

I didn't propose any changes to policy. I only suggested a change to the {{db-r3}} template so it more closely resembles the policy it's supposed to implement. To me the current text: This redirect may meet Wikipedia's criteria for speedy deletion as a recently created redirect from an implausible typo or misnomer, which is not in another language pertinent to the topic clearly implies that if the redirect is in another language that is not pertinent to the topic then it can be deleted under R3. You, however, say that it can't. One of these must be wrong. Warudo (talk) 19:13, 7 September 2025 (UTC)

The examples you gave are "in another language pertinent to the topic". They shouldn't have been tagged, as the tag already explains. WhatamIdoing (talk) 21:09, 7 September 2025 (UTC)

Just as a note, The Roman Catholic Archdiocese of Toronto was created as a redirect following reversion of an undiscussed, potentially controversial move. A category for Roman Catholic Archdiocese of Toronto exists, so it doesn't make sense for one to be name as the Roman Catholic Archdiocese of Toronto, in my opinion. You are free to bring Roman Catholic Archdiocese of Toronto to RFD or request the previous redirect be undeleted. Significa liberdade (she/her) (talk) 19:18, 7 September 2025 (UTC)

I don't care whether it exists or not, but it doesn't sound like you think The Roman Catholic Archdiocese of Toronto is "implausible". In fact, it sounds like you think it's so plausible that it could create confusion with the category name. WhatamIdoing (talk) 21:10, 7 September 2025 (UTC)

While not "typo", I could see "misnomer" being interpreted to include unrelated foreign-language names, particularly since the text specifically exempts related foreign-language names. I suppose the real test would be to find how often such nominations are actually made, and how often they go through versus are turned down or legitimately contested. Without being distracted by other (mis)uses of R3 like the The Roman Catholic Archdiocese of Toronto thing above. Anomie⚔ 15:09, 8 September 2025 (UTC)

I question whether we should judge the policy against 'less experienced editors sometimes nominate [and get declined and educated]' than against 'admins actually delete foreign language redirects under R3'. WhatamIdoing (talk) 15:57, 8 September 2025 (UTC)

@Anomie While I've called unrelated foreign language redirects very obscure synonyms from the perspective of an English speaker in an RfD once, I think calling them misnomers goes too far. They are the actual names used in some language after all. I also think that R3 was never intended to delete foreign language redirects at all. Originally, R3 did not mention foreign languages at all. Then, because people were misusing it and incorrect deletions were happening (see Wikipedia talk:Speedy deletion/Archive 15#Redirects in other languages and CSD R3) a complete prohibition of its use on all foreign language redirects (related or unrelated) was added. That was changed to the language we have today a few weeks later by a copyedit. Warudo (talk) 16:11, 8 September 2025 (UTC)

I'm not sure that linking to revisions from when it only said "implausible typos", before "or misnomers" was added to the criterion, is terribly useful as an argument. I also note that someone later added a significant word in there indicating that not all foreign-language redirects are useful. As I suggested, I think this is a case where looking at actual modern usage may be more useful than trying to find the original intent from 15 or 20 years ago. Or getting more people to actually weigh in. Anomie⚔ 16:45, 8 September 2025 (UTC)

Looking at modern usage does nothing to tell us if that usage is correct. We have to look at the actual discussions that formed the criterion to know what the correct interpretation is. Warudo (talk) 17:30, 8 September 2025 (UTC)

I agree that non-English redirects are only sometimes appropriate. However, a non-English redirect can be both inappropriate to have and also not appropriate to delete via CSD. The question isn't "Shall we keep non-English redirects?" but "Is the need for deletion so obvious in all cases that we can reliably skip the discussion at RFD?" WhatamIdoing (talk) 19:33, 8 September 2025 (UTC)

There's "correct" as in "what was written 20 years ago", and "correct" as in "the active community consensus being put into practice". You seem interested in the former. I'm more interested in the latter here. Anomie⚔ 20:58, 8 September 2025 (UTC)

I'm pretty sure deleting 4 year-old {{R from move}}s with R3 does not reflect consensus but it happens anyway (Memphis (daughter of Epaphus)). The active community consensus cannot be determined by looking at modern usage because admins use the tools contrary to consensus sometimes. To determine consensus we examine the actual text of the criterion. Contrary to what you may think, I'm not interested in what was written 20 years ago, I'm actually interested in correctly interpreting the criterion as its written right now. In my opinion, looking at the history of how the text became what it is today is helpful toward that goal. You may disagree, that's fine. Warudo (talk) 21:14, 8 September 2025 (UTC)

I think that looking at actual practice is an important way of determining consensus. Sometimes the written text in the policies and guidelines is outdated. That said, an occasional mistake or decision to Wikipedia:Ignore all rules shouldn't be treated like that mistake is the community's real practice today. WhatamIdoing (talk) 21:22, 8 September 2025 (UTC)

The problem I have with applying WP:EDITCON to speedy deletion is that only admins can delete pages. If deletion was an action that every editor could perform and revert, the deletion log would be indicative of consensus by definition, but it's not. To know what the community consensus is in this case, the way forward is getting more people to actually weigh in (as Anomie said above). Warudo (talk) 22:00, 8 September 2025 (UTC)

The three examples you've given show a somewhat inexperienced non-admin tagging your non-English redirects as R3. All three tags were promptly removed. None of them got deleted.

I looked through 10,000 (ten thousand!) deletions and found zero misuses of R3 to delete non-English redirects.

Therefore, I think it's safe to say that both admins and experienced editors agree that non-English redirects should not normally be deleted under R3.

What more can you want, than nothing getting deleted? Are you looking for "inexperienced editors will stop making mistakes"? That would be convenient, but it's not realistically available. WhatamIdoing (talk) 22:12, 8 September 2025 (UTC)

I guess you're right. For what it's worth, R3 is a criterion that is misused pretty often by people who don't follow the "recently created" requirement. Adding to the example above, I contested a few more such cases a month ago (and then got them deleted at RfD anyway). However, when it comes to the subject at hand, which is non-English redirects, misuse seems to be rare nowadays.

By the way, for anyone interested, I just remembered that WP:Proposal to revise CSD R3 is a relatively recent discussion about this stuff. Warudo (talk) 22:37, 8 September 2025 (UTC)

P.S: As it turns out, the option to add foreign languages to R3 got proposed and rejected last year anyway. I somehow missed this. Warudo (talk) 22:43, 8 September 2025 (UTC)

G5 clarity

• Om Murti Anil

Does CSD G5 apply if the majority of the content(around 75%, including the article’s creator) was added by socks from the same farm, all of whom are now blocked? Jeraxmoira🐉 (talk) 19:03, 5 September 2025 (UTC)

Once a speedy deletion has been declined in good faith (by someone other than the article creator) then it is not eligible for speedy deletion (there are some exceptions, e.g. for newly-discovered copyright violations, but they don't apply here), regardless of whether they meet the letter of a criterion or not. The reason for this is that speedy deletion is explicitly only for uncontroversial deletions, when someone objects it is not uncontroversial. Take it to AfD if you think it should be deleted. Thryduulf (talk) 20:11, 5 September 2025 (UTC)

The substantial edits by anonymous editors, not blocked as part of the sock farm, make this ineligible for G5. —David Eppstein (talk) 01:18, 6 September 2025 (UTC)

What number or % of substantial edits is required to retain the article or at what point does the % of contributions by socks justify deletion? Jeraxmoira🐉 (talk) 06:23, 6 September 2025 (UTC)

G5 demands "no substantial edits by others". None. —David Eppstein (talk) 06:28, 6 September 2025 (UTC)

G15 for files?

Should there be like a G15 equivalent for AI-generated images and videos without human review?

Like, this criteria would be like:

• A non-AI substitute exists or could be easily made
• The image does not have a legitimate encyclopedic use (e.g. the page on AI slop
• The image is very clearly AI-generated, with things like incomprehensible text. (similar to G15)

Anyone else think this would be a good idea? Lambda scorpii gmd (talk) 17:56, 7 September 2025 (UTC)

Does it meet the "frequent" criterion of proposing new criteria (see the top of this talk page)? —Alalch E. 18:29, 7 September 2025 (UTC)

There are no discussions currently transcluded at WP:FFD where being AI-generated is a concern. I also searched the FFD archives for AI generated and looked at every result, I found exactly zero instances of discussions where this criterion could have been used to delete an image where the discussion arrived at a consensus to delete. That's probably the most spectacular failure of the frequency requirement I've ever seen proposed. The second and third bullets of the proposal would also need workshopping to meet the objective and uncontestable requirements. Thryduulf (talk) 19:31, 7 September 2025 (UTC)

Okay I clearly did not do my research prior to suggesting that, I was just thinking like what if (which is probably a bad way to think). So sorry to waste you guys’ time. Lambda scorpii gmd (talk) 20:39, 7 September 2025 (UTC)

The main (and I think only good) reason to host images locally on en-wikipedia rather than on commons is that they fall under our fair use criteria. That is unlikely to be the case for most AI images, and if an AI image has a valid fair use rationale then that can only be because it does have a legitimate encyclopedic use. And this is not the place for discussing deletion criteria on commons, which has different standards and does host many AI-generated images. So I think there are, or should be, few or no images to which such a criterion could apply. —David Eppstein (talk) 20:54, 7 September 2025 (UTC)

That is unlikely to be the case for most AI images As a matter of fact, it's unlikely to be the case for any AI images, since AI images are allowed on Commons. JJP_Master (she/they) 13:36, 9 September 2025 (UTC)

Does A7 apply to academic journals?

A7 explicitly states that:

"it does not apply to articles about albums (these may be covered by CSD A9), products, books, films, TV programs, software, or other creative works, nor to entire species of animals."

On the other hand it says it covers "web content" so given that academic journals mostly nowadays are published online do they count as "books" or "web content"?

If not the latter, then what happens when a an article about a journal has no indication of importance? Xpander (talk) 14:10, 9 September 2025 (UTC)

It does not apply. If a journal has no indication of importance, conduct a WP:BEFORE search and then either expand the article if you find useful information, or WP:PROD or WP:AfD if you don't. Keep in mind that journals are one of those things that the community de facto maintains a pretty low notability bar for if they're reputable, even if there technically isn't a separate SNG for them. -- Tamzin^{[cetacean needed]} (they|xe|🤷) 14:15, 9 September 2025 (UTC)

(edit conflict) I'm pretty sure the reason that there are carveouts for albums/books/etc. is that in practice, a non-notable creative work is going to be a valid redirect title to the author (and WP:A9 is specifically to address the scenario where no article for a recording artist exists), and thus if not notable should almost always be WP:BLARed rather than deleted. Separately, I think in practice you'd have a hard time applying the "no indication of importance" speedy deletion criterion to a journal, as my sense is that even for very obscure journals, the claim of being a peer-reviewed is a plausible claim to importance (it may not be borne out, but that's a question for PROD/AfD, not CSD). _signed, Rosguill ^talk 14:16, 9 September 2025 (UTC)

If so shouldn't we add "academic journals", "magazine" along with books in the policy-line? Xpander (talk) 14:21, 9 September 2025 (UTC)

We could consider it and see if there's a consensus, but I'm mildly inclined to believe that it isn't really a problem that needs solving. As I laid out in my first comment, a page for a non-notable book is very likely to have a valid alternative to deletion. The same isn't quite true of magazines or journals that could be backed by an identifiable, notable institution, but could equally easily have a collective authorship that doesn't easily resolve to a single notable target (or alternatively, they could be published by an institution so noteworthy and famous and with so many publications that the one publication isn't WP:DUE for mention at the institution article. Contra Tamzin (a little), I think we could imagine an edge case for a self-described journal that does not claim to be peer-reviewed and does not present any claims of impact or importance, which could meet A7 criteria--but such cases are extremely rare _signed, Rosguill ^talk 14:27, 9 September 2025 (UTC)

A7 only applies to real people, individual animals, commercial or non-commercial organizations minus educational institutions, web content, and organized events, and as academic journals are neither of those things, A7 does not apply to academic journals. The list of what A7 is not is non-exhaustive and is entirely optional; it doesn't affect the policy norm at all and is just built-in commentary. —Alalch E. 15:15, 9 September 2025 (UTC)

@Rosguill, How does the fact of being peer-reviewed (which is a norm for academic journals) count towards a journal's notability? It's not mentioned on Wikipedia:NJOURNALS. Furthermore how is such a claim even verified? Any journal could claim they are peer-reviewed, but how do we establish the fact that it is (except for well-established journals)? Xpander (talk) 05:52, 10 September 2025 (UTC)

Remember that a claim of significance is explicitly a lower standard than notability, and doesn't have to be correct, let alone verified as correct, just plausible. Being peer reviewed is a claim of significance. Thryduulf (talk) 11:01, 10 September 2025 (UTC)

Xpander, you can establish this by asking someone with Ulrichsweb access. We librarians use this massive database when making professional decisions about individual serials (it has an extremely high reputation for accuracy), and because it's relevant to scholars publishing in specific journals and scholars researching the significance of individual serials (whether journals or not), far more people than librarians would have access to it. I suspect most anglophone universities would have it, and WP:TWL or WP:RX might be able to help, too. Nyttend (talk) 20:47, 26 September 2025 (UTC)

Does A7 apply to historical places?

This a follow up to the discussion above (on academic journals), does A7 apply to historical places? If not what would be the criterion for speedy deleting those pages? Xpander (talk) 05:44, 10 September 2025 (UTC)

Not according to the rubric: A7 ... applies to any article about a real person, individual animal, commercial or non-commercial organization, web content, or organized event that does not indicate why its subject is important or significant, with the exception of educational institutions. A historical place is not a person, an animal, an organization, web content or an event. Cheers, SunloungerFrog (talk) 06:08, 10 September 2025 (UTC)

Thanks. Xpander (talk) 08:23, 10 September 2025 (UTC)

New speedy deletion criteria "G16. Empty talk pages"

The following discussion is closed. Please do not modify it. Subsequent comments should be made on the appropriate discussion page. No further edits should be made to this discussion.

---

This will apply to:

• empty talk pages
• talk pages that have no subjects and just have notices or templates

but it will not apply to:

• talk pages that are empty/just have notices or templates because of archiving
• user ones

BodhiHarp 18:42, 10 September 2025 (UTC)

What is the problem that you are trying to solve here, and can it not be solved with existing processes? 331dot (talk) 18:45, 10 September 2025 (UTC)

Before replying: Edit: It also won't apply to talk pages rating the article etc. BodhiHarp 18:48, 10 September 2025 (UTC)

@331dot Wouldn't this simplify having to discuss deleting it? BodhiHarp 18:50, 10 September 2025 (UTC)

I guess what I'm getting at is, why does an empty talk page need to be deleted at all?

We already have G2(any page created as a test) and G8(pages associated with a nonexistent page) which sounds like covers many blank talk pages. What's left? 331dot (talk) 18:55, 10 September 2025 (UTC)

G2 could maybe cover it. BodhiHarp 19:00, 10 September 2025 (UTC)

Which leaves the question- why is this needed? I think doing as Whpq asks will help answer this. 331dot (talk) 19:03, 10 September 2025 (UTC)

(edit conflict) If the only content in the history is nonsense or vandalism (and blanking to remove it) rather than test edits, G1 or G3 could apply. Anomie⚔ 19:05, 10 September 2025 (UTC)

Considering what happened in the "New speedy deletion criteria" discussion above, did you go through the 4 points listed at the top of the talk page? If so can you provide your analysis of how each point is met by your proposal? -- Whpq (talk) 19:01, 10 September 2025 (UTC)

I do now realize that other speedy deletion criteria could be used, but what about the case where it's not vandalism, but the user simply didn't know? Maybe perhaps it could be hard to tell. BodhiHarp 01:55, 11 September 2025 (UTC)

Again, how does your proposal meet the 4 criteria? -- Whpq (talk) 01:58, 11 September 2025 (UTC)

Maybe it might not actually. BodhiHarp 02:02, 11 September 2025 (UTC)

Editors proposing a new CSD need to think through the proposal. I suggest you just ask this to be closed now. -- Whpq (talk) 02:26, 11 September 2025 (UTC)

The above is preserved as an archive of the discussion. Please do not modify it. Subsequent comments should be made on an appropriate discussion page (such as the redirect's talk page or in a deletion review).

Why is "G6: Housekeeping and routine (non-controversial) cleanup" still in the delete dropdown?

Yes, "because no one has changed MediaWiki:Deletereason-dropdown and MediaWiki:Filedelete-reason-dropdown". Why haven't they?

I was curious as to when G6 changed from "Housekeeping" to "Technical deletions", particularly given how much some people complain about G6 being misused to the point where Wikipedia:What G6 is not is "See also"-ed at the top of the criterion and yet "Housekeeping and routine (non-controversial) cleanup" is so attractive for such misuse if someone hasn't carefully read WP:CSD lately. I was surprised to find out that change happened way back in June 2008. If you really believe what that essay says, why hasn't it long since been replaced with specific entries for "G6: Deletion to make way for a page move", "G6: Unambiguously created in error or in the incorrect namespace", "G6: Redirect created by moving away from a title that was obviously unintended", and "G6: Template orphaned after TFD"? Anomie⚔ 19:36, 10 September 2025 (UTC)

Probably because nobody's gotten around to it. You're an admin, so do feel free to. Stifle (talk) 12:40, 12 September 2025 (UTC)

I wanted to ask first if there was a reason no other admin had done so, and to see if anyone wanted to oppose. If no one has by sometime next week, I'll probably go ahead. Anomie⚔ 13:49, 12 September 2025 (UTC)

Made the edits. We'll see if anyone notices. Anomie⚔ 13:11, 17 September 2025 (UTC)

Noticed the change. I'd still include a general G6 option on the menu, just changed to "Uncontroversial technical maintenance" or "Technical maintenance" to give a clue that some technical reason is needed beyond "nobody would care, not needed". I'd also update {{db-g6}} to use the same summary and change the error dropdown to just "Obviously created in error" like the {{db-error}} summary, as the wrong namespace is still an error. ~ Jenson (SilverLocust 💬) 07:34, 23 September 2025 (UTC)

I'm not sure that either "Uncontroversial technical maintenance" or "Technical maintenance" really signals that a reason needs to be provided. But I won't try to stop you from making that change.

Given that WP:CSD#G6 contains "or in the incorrect namespace", I'd leave that in the dropdown. There may be someone around who considers "in error" to strictly mean "shouldn't have been created at all", excluding "creation was ok but in the wrong namespace". Anomie⚔ 13:41, 23 September 2025 (UTC)

I also noticed the change. There are way more than just two specific reasons for G6. If you remove the generic housekeeping option, I'll just delete without specifying any speedy criterion. I usually specify my reason in detail anyways on most of my G6 deletions. Talk:Homo superior. – wbm1058 (talk) 12:29, 26 September 2025 (UTC)

Yes, I've also just noticed it. Many of the deletions I do under this flag are not the two options that have been left behind, for example when conducting history merges. Housekeeping should be just that, uncontroversial maintenance that doesn't fall under the other specific deletion reasons, and I have no idea why we'd remove this tool from admins when it's just going to lead to less clarity. Anomie please revert your changes unless and until there is a much wider consensus for them. (We can keep the more specific instances of G6 if you like, but the general housekeeping one should still be present)  — Amakuru (talk) 10:42, 29 September 2025 (UTC)

Personally I had just gotten a bit tired of the mismatch between what the people who maintain this policy say in Wikipedia:What G6 is not (which they've even linked from the top of WP:CSD#G6) and the messaging in the interface which very much didn't match it. You're an admin too, you can just as well improve the messages, but I'd recommend you don't just revert the change as that would reintroduce the mismatch between the interface and policy. Even just reintroducing the "housekeeping" language wouldn't really match the current policy; you'd probably do better to say "technical deletions" instead to match what G6 has been titled as since 2008.

BTW, there are four specific reasons now included, not just two: "for a page move" and "in error" in the "general" section, "redirect from unintended title" in the "redirect" section, and "template orphaned after TFD" in the "template space" section (which only shows in the Template and Module namespaces). There's also nothing stopping you from typing [[WP:CSD#G6|G6]]: Whatever reason you'd like into the form without using the dropdown at all. Anomie⚔ 12:49, 29 September 2025 (UTC)

Per the discussion above, I have now added [[WP:CSD#G6|G6]]: Technical deletion (uncontroversial maintenance) [44]. This wording aligns with what the current text of WP:G6 says: G6. Technical deletions. This is for uncontroversial maintenance, including: .... Mz7 (talk) 23:28, 29 September 2025 (UTC)

Also, not that I necessarily disagree with the content of the essay, but I just wanted to note that WP:What G6 is not only became linked as a hatnote to WP:G6 in January 2024 as a presumably WP:BOLD edit [45]. I'm not sure the extent to which that essay has been vetted by the community. I know that historically some admins have been more cavalier with G6 than others, and I'm not sure if we've had any recent community-wide discussions that have explicitly endorsed the view in the essay that G6 should rarely be used to delete pages that do not explicitly fall into one of the above bullet points. Mz7 (talk) 23:34, 29 September 2025 (UTC)

I think we should abolish G6 entirely and split deletions to make way for moves and deletions of pages obviously created in error into their own separate criteria (and get rid of Deleting templates orphaned as the result of a consensus at Wikipedia:Templates for discussion. entirely; that's not a speedy deletion at all, it's a deletion per a deletion discussion). * Pppery * _{it has begun...} 23:52, 29 September 2025 (UTC)

Agreed. G6 would never gain consensus if proposed today as it lumps together multiple poorly defined and only tangentially-related things. There should be separate criteria for:

• Temporary deletions (e.g. for history merges)
• Pages created in error and redirects created while fixing such errors
• Pages in the way of a page move
• Pages created temporarily as part of a page move
• Pages temporarily undeleted (e.g. for a deletion review)
• Anything else that meets the normal NEWCSD requirements.

Thryduulf (talk) 01:12, 30 September 2025 (UTC)

Yeah, it's a bit of a mess now. Stifle (talk) 12:59, 30 September 2025 (UTC)

Can redirects to removed/renamed sections be G8'd?

I came across Wikipedia:Redirects for discussion/Log/2025 September 12#Rotarian misogyny, an RfD that was closed as speedy delete via G8. The deleted redirect was an {{R to section}} and the target section has been removed due to NPOV violations here. By the literal letter of G8 this is wrong since G8 deletes [r]edirects to target pages that never existed or were deleted (emphasis mine), but since G8 uses the phrase [e]xamples include, but are not limited to which makes it open-ended, one could decide to extend the criterion to [r]edirects to targets that never existed or were deleted which would arguably make the deletion correct. Warudo (talk) 15:29, 13 September 2025 (UTC)

I would say in most cases no. G8 is a broad criterion, and the list of use cases it gives is explicitly non-exhaustive, so I don't want to entirely preclude the possibility there could be a case where a redirect is truly "dependent" on the section it redirects to. But usually that won't be the case, because redirects without mention are not categorically prohibited (even if they're rather disfavored), and there may be relevant content elsewhere in the article or on another page. -- Tamzin^{[cetacean needed]} (they|xe|🤷) 16:14, 13 September 2025 (UTC)

Definitely not. The only time G8 should be used with a redirect is if the target page does not exist (i.e. it is currently a redlink) and the criterion really ought to be updated to reflect this. Sections get renamed, moved, merged and split all the time so the target section not existing means does not reliably imply anything about what should happen to the redirect (i.e. expanding G8 to included redirects to sections of pages that do exist would be a gratuitous failure of NEWCSD point 2). In the linked case, speedy deletion as vandalism or an attack page might have been justifiable (I haven't looked in enough detail to be certain) but G8 was unambiguously incorrect. Thryduulf (talk) 16:56, 13 September 2025 (UTC)

G8 obviously does not apply in this case. Even if its listed criteria are not exhaustive, it is only for "pages dependent on a non-existent or deleted page". Sections are not pages, and removing them from their articles does not automatically make {{R to section}}s useless, as they can be retargeted to other valid articles and sections if they exist (Sexism on Rotary International (not "Misogyny on...") would have been a marginally acceptable redirect to this section). If it were not for the fact the term "Rotarian misogyny" doesn't appear anywhere else, I would have considered raising objections to the speedy. Dsuke1998AEOS (talk) 19:27, 13 September 2025 (UTC)

Let's imagine we added this criterion right now and consequently deleted thousands of redirects: Most of those deletions would be really bad and much worse than doing nothing.—Alalch E. 00:36, 14 September 2025 (UTC)

Most certainly not. In this situation, G8 is strictly for redirects that outright do not work — they lead you to a page that produces MediaWiki:noarticletext. R-to-section still takes you to the target article, and whether it's acceptable without the section in question must be judged by consensus, not by speedy deletion. Nyttend (talk) 20:41, 26 September 2025 (UTC)

Since we have a bunch of others weighing in here: IMO Tamzin has it right, most likely the redirect can be targeted to a different section or to the article as a whole rather than being deleted. And I particularly note that the absolutist "G8 is 100% only for broken redirects, there's never anything else that 'dependent on' could mean" position has proven to be a small minority in other contexts. Anomie⚔ 21:54, 26 September 2025 (UTC)

Why are U1 and G7 separate criteria?

Why should these be separate. BodhiHarp 22:09, 13 September 2025 (UTC)

It is correct that in the vast majority of cases where U1 is applied, G7 could be used instead, but the criteria are different: U1 applies to all pages in a user's userspace, even to those with major contributions by other people. G7 only applies if there are no significant contributions by others. Merging the two is probably possible, but I don't see any major benefits. —Kusma (talk) 22:19, 13 September 2025 (UTC)

(edit conflict) Because U1 allows someone to delete a page in their own userspace, even if they are not the only contributor to the page. G7 applies to all namespaces but only when the nominating editor is the sole significant contributor. They overlap, but not enough that one is a subset of the other. Combining them would lead to a more complicated criterion, which is generally less than ideal - simple criteria are easier to remember and interpret and are less likely to be used incorrectly (intentionally or accidentally). Thryduulf (talk) 22:23, 13 September 2025 (UTC)

Another key difference is U1 applies even if there was a previous XfD. HouseBlaster (talk • he/they) 23:39, 13 September 2025 (UTC)

RfC on G14 for redirects

The following discussion is closed. Please do not modify it. Subsequent comments should be made on the appropriate discussion page. No further edits should be made to this discussion.

---

The WP:G14 criterion for Redirects that end in "(disambiguation)" but target neither disambiguation pages nor pages that perform disambiguation-like functions (such as set index articles or lists), could apply where the term in the redirect title is not the ambiguous term disambiguated at the target. Shhhnotsoloud (talk) 11:18, 14 September 2025 (UTC)

• Support as nominator. See, for example, Wikipedia:Redirects for discussion/Log/2025 September 13#Lists of Japanese Governors-General (disambiguation). If Seventeen ducks (disambiguation) were to redirect to Lists of Japanese governors-general then it could be G14'd because Lists of Japanese governors-general doesn't disambiguate "Seventeen ducks". Pertinent to the particular example, Lists of Japanese governors-general (plural) could also be G14'd because Lists of Japanese governors-general doesn't disambiguate "Lists of Japanese governors-general" (which is not ambiguous). Shhhnotsoloud (talk) 11:24, 14 September 2025 (UTC)
• Strong oppose. Whether something is related to the target is often subjective, and much better determined through discussion than speedy deletion. This is also not a frequent occurrence at RfD so I highly doubt WP:NEWCSD's frequency requirement is met. Finally, was there really a need to jump straight to an RFC about this? Where have you discussed your proposed change to policy previously? Thryduulf (talk) 11:29, 14 September 2025 (UTC)
  • I am following the suggestion in the discussion thread in which you participated at RfD where "three people ... are telling you that your interpretation of policy is wrong", so help clarify the policy, don't criticise me for asking for it to be clarified. Shhhnotsoloud (talk) 11:42, 14 September 2025 (UTC)

    The advice was to discuss your interpretation of G14 here if you wanted more opinions than just us three, not to start an RFC to change the policy without prior discussion about the change. Thryduulf (talk) 13:54, 14 September 2025 (UTC)

• Oppose. I had to read the proposal several times to understand what you mean. This seems to add complexity to G14 without any evidence of need. —Kusma (talk) 12:23, 14 September 2025 (UTC)

  Me, too. And now that I think I understand it, I also oppose it. WhatamIdoing (talk) 22:45, 24 September 2025 (UTC)

• I'm not totally sure I understand the question here. If it's whether "Foo (disambiguation)", pointing to "Bar", a DAB(-like) page that doesn't disambiguate "Foo", should be eligible for G14, my tentative answer is no, on the basis that whether the article disambiguates the term can be subjective. If such a redirect winds up existing despite literally no connection to the target page, any of G3, G6 (as error), or R3 might apply. -- Tamzin^{[cetacean needed]} (they|xe|🤷) 12:32, 14 September 2025 (UTC)

  If [the question is] whether "Foo (disambiguation)", pointing to "Bar", a DAB(-like) page that doesn't disambiguate "Foo", should be eligible for G14 that is my understanding of the question. Thryduulf (talk) 14:19, 14 September 2025 (UTC)

  Yes, that is what I meant. Shhhnotsoloud (talk) 17:39, 14 September 2025 (UTC)

  Why would you delete that, instead of just repointing the redirect (e.g., make "Foo (disambiguation)" point to the Foo article)? That seems entirely WP:SURMOUNTABLE, at least in many cases, and therefore ineligible for CSD. WhatamIdoing (talk) 22:47, 24 September 2025 (UTC)

• Oppose. Not sufficiently objective and not frequent enough from what I can tell. I also agree that it adds an amount of complexity to the criterion that is a detriment not in proportion to the potential benefit.—Alalch E. 17:44, 14 September 2025 (UTC)
• Oppose - Not needed as already stated by others. -- Whpq (talk) 17:46, 14 September 2025 (UTC)
• Oppose. I read the proposal three times and still don't understand what it means. It therefore needs to be reworded and simplified. Stifle (talk) 07:59, 18 September 2025 (UTC)
• Oppose Counter Stifle I do understand what is being proposed here, but I think this isn't common enough to warrant speedy deletion. * Pppery * _{it has begun...} 15:51, 20 September 2025 (UTC)

The discussion above is closed. Please do not modify it. Subsequent comments should be made on the appropriate discussion page. No further edits should be made to this discussion.

Definition of "recently created"

@Green Montanan Isn't 1000 a bit too high? The TFA as of writing this, Caerleon pipe burial needed 2 months to get 1000 views. 6 weeks after its creation it had ~500. In my opinion, even 500 views are still too many for A10 or R3. Warudo (talk) 15:16, 21 September 2025 (UTC)

For anyone reading this who wants context, I'm talking about this edit. Warudo (talk) 15:23, 21 September 2025 (UTC)

Remember: it's 1,000 views over the span of 3-6 weeks, which translates to an average of 24 to 48 views a day (for 6 weeks and 3 weeks respectively).

But I don't really care what number we use. It's just that unlike the intentionally vague term "recent", I don't see the benefit of being vague about pageviews. Green Montanan (talk) 16:01, 21 September 2025 (UTC)

We really shouldn't use a number. "High-profile" isn't just about page views, so I will revert to the previous version. —Kusma (talk) 16:37, 21 September 2025 (UTC)

How else would you measure "high-profile"? Green Montanan (talk) 18:43, 21 September 2025 (UTC)

Or phrased another way, what else is there to a page being "high-profile"? Green Montanan (talk) 18:45, 21 September 2025 (UTC)

@Kusma Did you also intend to revert Special:Diff/1312575341? I don't think that was a bad change and you didn't mention it in your edit summary. Warudo (talk) 20:47, 21 September 2025 (UTC)

@Warudo, I think @SilverLocust's edit summary here clarifies that page moves are not the only thing to be considered here. —Kusma (talk) 16:22, 22 September 2025 (UTC)

R3 is already one of our weirdest criteria. As written, it'll always be more of an "I know it when I see it" than a bright line. There's two obvious alternatives to that—abolish it, or define it much more strictly—but the community's never seemed much interested in that, I think because the "I know it when I see it" approach, anathema to WP:NEWCSD as it may be, in practice works pretty well. Admins delete weird redirects that'll never survive RfD, and in the rare cases where someone challenges that, RfD tends to affirm the admins' judgment. Given all that, I'm not sure how important it is to define this one term that only appears in a footnote meant to explain another vague term, as far as R3 is concerned. It may be more relevant for A10 but I don't know; is there an indication that the lack of an exact number is causing problems? -- Tamzin^{[cetacean needed]} (they|xe|🤷) 16:07, 21 September 2025 (UTC)

Why wait for problems to arise? Why not proactively eliminate problems? Green Montanan (talk) 18:46, 21 September 2025 (UTC)

Its process for process sake, thats why. Spartaz ^Humbug! 18:52, 21 September 2025 (UTC)

What does that mean? Green Montanan (talk) 19:09, 21 September 2025 (UTC)

"Process for process sake" means making up rules today, just in case they might be needed at some point in the future. WhatamIdoing (talk) 22:49, 24 September 2025 (UTC)

Is that bad thing? Why wait for the horse to run away before putting a lock on the barn door? Green Montanan (talk) 15:33, 26 September 2025 (UTC)

Why put a lock on the barn door, when there is no horse worth stealing? WhatamIdoing (talk) 15:53, 26 September 2025 (UTC)

I don't really agree with this take. R3 is misused a lot, especially by editors who ignore the "recently created" rule and delete redirects that are several years old. For example, Memphis (daughter of Epaphus) was 4 years old (actually 7 years old since it was an R from move from even earlier) when it was deleted with R3. Sure, I don't think it would survive RfD so I didn't contest the deletion but it was still wrong to use R3. At best it's an IAR deletion. Perhaps we should consider either a) abolishing R3, b) rigorously defining "recently created" with a bright-line rule or c) getting rid of "recently created" entirely.

On a related note, I recently contested an R3 where an admin considered an {{R from sort name}} redirect implausible enough to R3. That one is at RfD as of writing this so it may end up deleted anyway. Still, I believe that the R3 was wrong because the existence of the rcat shows that sort names are not considered implausible. In short, R3 does not work pretty well, it's just that redirects aren't that visible so people don't notice when they are incorrectly deleted. Warudo (talk) 12:09, 22 September 2025 (UTC)

If an administrator is interpreting a four-year-old redirect as being "recently created" (let alone seven!), then they're not going to be deterred by an explicit numeric limit either. The problem isn't the criterion. The problem is the people. —Cryptic 16:37, 22 September 2025 (UTC)

Unless we write an edit filter warning admins about too old deletions, which would be easier to do with an objective definition. I actually proposed something like this at WP:EFR once and it didn't go anywhere. * Pppery * _{it has begun...} 17:21, 22 September 2025 (UTC)

I'd like to see Twinkle notify taggers of problems like this. See Wikipedia talk:Twinkle#Double checking feature. WhatamIdoing (talk) 22:50, 24 September 2025 (UTC)

FYI I also noticed Memphis (daughter of Epaphus) being deleted out of process at WP:Database reports/Possibly out-of-process deletions#Not recently created (various criteria) (which uses 4 months), and also decided not do do anything about it for the same reason as Warudo. * Pppery * _{it has begun...} 17:21, 22 September 2025 (UTC)

I don't necessarily mind putting some numbers on this, but this particular number is off by a couple of orders of magnitude.

Looking at the table in WP:VIEWSSTATS, only 20% of Wikipedia articles get 1,000 page views in an entire year. It doesn't make any sense to say that "very few page views" is more page views than 80% of our articles will get in a whole year, especially when you remember that they're supposed to have achieved this feat of popularity in (at most) "about 3–4 months".

The median Wikipedia article gets a total of about 20–30 page views in 3–4 months. I don't think that the median is "very few page views"; that's better described as "the median page views" or "the normal number of page views". It drops off steeply from there, though: a solid third of our least popular articles get no more than three (3) page views in 3–4 months.

But: newly created pages (including redirects) get checked by Wikipedia editors, and that process alone can generate a one-time bump of 25–50 page views (though probably not so many for redirects), especially within the first 24 hours.

Perhaps we could use these stats to find a number that is appropriate. I'd suggest maybe "averaging of one per day" as being high enough that it can no longer be considered "very few page views". WhatamIdoing (talk) 16:06, 26 September 2025 (UTC)

Requesting admin review of G6-tagged article (duplicate of existing draft)

I just recently tagged Rico Salmela for deletion under under G6. A complete, well-sourced draft already existed (since September 6) at Draft:Rico Salmela, created by me, which was submitted for review yesterday and was waiting in queue. The current mainspace article was created after my draft (just ignored the draft's existence completely), lacks depth, contains incorrect information and duplicates the topic of the draft. I’m requesting deletion so the draft can be moved into mainspace instead.

Requesting this here for visibility as the tag on the article hasn't been acted on yet. Thanks. Fishyblyn (talk) 09:19, 23 September 2025 (UTC)

@Fishyblyn: This isn't the right venue for requests about specific pages, but, regardless, I've declined G6 because that is not one of the valid uses of G6 (and wouldn't be a valid reason for any kind of IAR deletion either). People are allowed to create an article while a draft is being worked on. You can now improve the article by manually merging in what you have in the draft. -- Tamzin^{[cetacean needed]} (they|xe|🤷) 09:29, 23 September 2025 (UTC)

I have improved the article by doing just that. The creator and sole author had requested deletion in the meantime, and the draft had been deleted, but I requested undeletion from the deleting admin, and he undeleted, and so I performed the merger, providing attribution. —Alalch E. 13:45, 23 September 2025 (UTC)

New speedy deletion criteria "C5. Categories populated by a single page"

I propose we add this new criterion, which, like its title says, are categories populated by a single page. However, it does not apply to categories that are currently populated with a single page, but there are other pages that fall in the category and should have the category. Yes, I read the page and in summary it has a potential and probably should be discussed.

1. Objective: ✓ Potentially possible to tell whether a page meets this, but we might need to discuss it
2. Uncontestable: X Need to discuss this
3. Frequent: X Need to find out how frequently this occurs, maybe not
4. Nonredundant: X Maybe we should use this instead of WP:CFD, but we should discuss whether it is frequent and uncontestable

BodhiHarp 02:28, 30 September 2025 (UTC)

So, to be clear: You think it would be preferable to delete, say, Category:Ivorian women mathematicians and have an inconsistent category structure than to keep it and have a consistent structure with some one-element categories? That seems...not uncontroversial to me. Speedy deletion should be for things that are uncontroversial. —David Eppstein (talk) 02:32, 30 September 2025 (UTC)

CfD sometimes does delete pages populated by a single page, but it also sometimes doesn't. Example with no consensus at Wikipedia:Categories_for_discussion/Log/2025_September_3#Category:Water_parks_in_Austria. That means this isn't uncontestable. (And a minor technical sidenote: this would be better addressed through the CFDS process than speedy deletions even if there were a consensus to do it.) * Pppery * _{it has begun...} 02:37, 30 September 2025 (UTC)

This isn't a new criterion, it's an amendment to C1. —Cryptic 04:29, 30 September 2025 (UTC)

• As you have yourself acknowledged, this doesn't seems to meet the uncontestable and frequent criteria. Oppose. Stifle (talk) 12:56, 30 September 2025 (UTC)

RfC: Replacing U5 with a primarily procedural mechanism

Should CSD U5 be 1) repealed and replaced with a combination of 2) procedural deletion of non-contributors' user subpages after six months of no edits, 3) a narrower criterion for off-topic content that has escaped deletion under (2), 4) formalizing the practice of moving drafts off of top-level userpages, and 5) allowing editors to blank userpages that would be eligible for speedy deletion under (3) if they were subpages? -- Tamzin^{[cetacean needed]} (they|xe|🤷) 10:59, 30 September 2025 (UTC)

The exact details of this proposal are:

1. Repeal CSD U5.
2. Enact CSD U6, "Abandoned user subpages of non-contributors":

   User subpages of users who have made few or no edits outside of user space, which have not been edited by a human in at least six months. Promising drafts may be moved to draftspace by any editor as an alternative to deletion.

3. Enact CSD U7, "Excessively unrelated or grossly improper non-draft subpages by non-contributors":

   User subpages of users who have made few or no edits outside of user space, which were created more than six months ago, could not be interpreted as draft articles (even very bad ones), and unambiguously violate the userpage guideline's sections "Excessive unrelated content" or "Advocacy or support of grossly improper behaviors with no project benefit". See WP:UPNOT regarding handling of similar material on top-level user pages.

4. In WP:UPYES, remove "usually" from Work in progress or material that you may come back to in future (usually on subpages) and add a third sub-bullet to it:

   Drafting on a top-level userpage is confusing. Any editor may move a draft away from a top-level userpage, either to a subpage or to the draft namespace, replacing it with {{draftified userpage}}. If the user page's owner reverts such a move, this is taken as a statement that their userpage should not be viewed as a draft for the purposes of assessing compliance with this guideline, which may make it qualify as excessive unrelated content.

5. In WP:UPNOT, append to third paragraph:

   For users with few or no edits outside user space, excessively unrelated or grossly improper subpages more than six months old may be deleted under speedy deletion criterion U7; if a top-level userpages would be eligible for deletion under that criterion if it were a subpage, it may be blanked by any editor. Note that this excludes drafts, which should be handled as described above.

-- Tamzin^{[cetacean needed]} (they|xe|🤷) 10:59, 30 September 2025 (UTC), small copy-edits 14:11, 30 September 2025 (UTC)

References

Survey

• Yes. The status quo of U5 is the worst of both worlds. On the one hand, its vagueness and subjectivity leave editors far too willing to tag and delete a wide variety of valid userspace pages, including full-fledged drafts and innocent self-introductions. In my experience as a patrolling admin, as currently stands the average U5 nomination is against policy (although often still deletable under other criteria, especially G11). On the other hand, it sucks up patroller and admin time leafing through pages that are largely harmless individually but dangerous in aggregate, while still missing lots of instances of that that just fall through the cracks. Few editors police old userspace subpages, meaning that even if a page could be deleted under a G-series criterion, if it survives tagging in the first few days after creation it may well never be dealt with, even if it contains BLP violations, SEO spam, etc. By having no procedural means of deletion in userspace, but relying on a vague catch-all I-don't-like-it criterion, we essentially guarantee negative interactions for potential constructive new users, while hosting a large enough swath of unmaintained userspace content that it's very difficult to catch actual bad actors who didn't get caught immediately.
  The logical solution to this is to make the deletion of unmaintained pages in non-contributors' userspace procedural, the same as it is for unmaintained drafts in draftspace. This means that the vast majority of U5 cruft will be deleted without anyone needing to assess it on the merits. U5 content is not particularly time-sensitive to delete; anything that needs to be deleted now, not six months from now, ought to be deletable under another criterion like G11 or G3. The two loopholes this would leave are where a subpage is ineligible due to being actively maintained but is completely unsuitable for Wikipedia, which the proposed U7 (preserving a stricter version of the existing U5 logic) is there to address, and where the content is on a top-level userpage, which the proposed UPNOT/UPYES changes are there for. If top-level userpage content doesn't fall under any other criterion, there shouldn't be any harm in leaving it in a page history; and if a user repeatedly restores content to their userpage that is clearly unsuitable, well, that's something they're already able to do by repeatedly recreating a deleted userpage, and in both cases this becomes a user conduct issue suitable for AIV or AN/I. The remaining cases within the current U5 scope would be borderline calls, like a page that is ambiguously either a draft or an overlong curriculum vitae, but is being maintained and so won't get deleted procedurally; those cases are inherently subjective and belong at MfD.
  The other concern is, of course, deletion of useful content. I don't think this proposal would increase the rate of that, and if anything will decrease it. Currently, in the rare cases where a non-contributor writes a viable userspace draft and then abandons it, the default outcome is that it stagnates forever. Under this system, there's actually a mechanism for someone to eventually put eyes on the page and potentially move the page to draftspace, where it's moderately more likely to be noticed. When that doesn't happen, there would always be REFUND; we could rename WP:REFUND/G13 to WP:REFUND/G13 and U6. -- Tamzin^{[cetacean needed]} (they|xe|🤷) 10:59, 30 September 2025 (UTC)
• Against enacting CSD U6, support CSD U7, wp:upyes change and wp:upnot change. The replacement of CSD U5 should allow the user to experiment until they feel like they are ready to create a draft. Failing to allow that would stifle the growth of new users, which is not helpful in the long term for wikipedia. Snævar (talk) 12:37, 30 September 2025 (UTC)
  • I'm confused. U6 (unlike U7) does let the user experiment as long as they like - it's based on the time of the most recent edit, not the creating one. Even if they take a six-month break mid-experiment and the page is deleted, they can get it back procedurally at WP:REFUND, and I expect the standard deletion log text to point there the same way G13's does. And it's worlds better than what we have with U5-as-practiced. —Cryptic 14:00, 30 September 2025 (UTC)
• I can get on board with this. U5 doesn't meet the objective criterion anymore, and I'm not sure it ever did. The rest of the changes just follow obviously. Stifle (talk) 12:58, 30 September 2025 (UTC)
• Yes. From my past patrolling of the user namespace from recent changes and participation at MfD and, I have come to realize that many or most editors do not have a good concept of whether someone is a contributor or a non-contributor. The concept of non-contributor is just bad for human application, causing U5 to be a bad criterion for human application as a whole. Many, for example, do not understand that a completely new user creating a draft automatically and very strongly marks that user as a contributor (they were not a contributor because they had not been a Wikipedia user, then they became a user and immediately presented themselves as a contributing user, i.e. contributor). They will then tag this brand new contributor's draft for deletion instead of moving it to draftspace or similar. That user might've been right here and could have been welcomed, notified, talked to. Instead, the very first interaction they might get is an improper tagging of their first page for deletion. At the same time, many userspace creations go unseen (they are not very interesting to patrol, unless you have a morbid curiosity), leaving the actually bad pages, of which there are many, indefinitely hosted. They need to be deleted using a catch-all criterion.—Alalch E. 13:18, 30 September 2025 (UTC)
• What: Unqualified yes on #1 (repeal U5), #2 (U6: G13-like timeout on subpages), and #4 (UPYES change). Very reluctant ok on #3 (U7: old unrelated content) based on its wording rather than its idea (see discussion below), and its dependent change #5, since while flawed, they're better than what we have and a fill a hole that U6 by itself would leave.
  Why: my specific previous comments here about the status quo as practiced and here about the impact on new users. —Cryptic 14:10, 30 September 2025 (UTC)

Discussion

Wikipedia talk:User pages notified. Courtesy pings to Wikipedia:Village pump (idea lab) § Rethinking CSD U5 participants @Cryptic, WhatamIdoing, Chaotic Enby, Thryduulf, asilvering, Perryprog, Anomie, Pppery, rsjaffe, Vanamonde93, Chipmunkdavis, Clovermoss, MZMcBride, Risker, and Stevenarntson; will manually notify 184.152.65.118. There were a lot of great ideas in the VPIL thread, many of which I've incorporated, but some of which I didn't because they were too far-removed from the core proposal or irreconcilable with other ideas. But I've tried to frame this in a somewhat severable way, such that e.g. someone could !vote "support except for U7" if they wanted. -- Tamzin^{[cetacean needed]} (they|xe|🤷) 10:59, 30 September 2025 (UTC)

• Comment: The proposal is quite clear about what is suggested should be done. But we seem to lack the reasons for the proposal. Obviously, by having a proposal to change a rule in Wikipedia, as proposed here, something is wrong or not working properly, etc. But what? It might seem clear to the author of this proposal but, methinks, they should offer a detailed reasoning, hopefully with some indicative examples, before we start digging in. Cheers. -The Gnome (talk) 11:59, 30 September 2025 (UTC)

  @The Gnome: See my comment above. If you want illustrations of how often U5 is misused, a few years ago I posted this analysis of a set of deletions. You can ctrl+f "decline U5" in my userspace contribs to get a sense of the kinds of things that get incorrectly tagged, although what that doesn't show all the times that the U5 tag is invalid but it is G11 or something else. At least one that I've declined on has since made it through NPP as an article. -- Tamzin^{[cetacean needed]} (they|xe|🤷) 12:22, 30 September 2025 (UTC)

• One minor nitpick, I'd change few or no edits outside of user pages to few or no edits outside of user space. That makes it less likely someone could misinterpret the following "which have not been edited" to be attached to the no-edits clause rather than the subpages being deleted, and makes it less likely for someone to waste time arguing that user-talk page edits count as "edits outside of user pages". Anomie⚔ 13:16, 30 September 2025 (UTC)

   Done -- Tamzin^{[cetacean needed]} (they|xe|🤷) 14:13, 30 September 2025 (UTC)

• These are clear improvements over what we have (which in practice is "Any page the tagger and deleting admin don't like, in the userspace of someone too new to effectively complain"). However - and I'm sorry I didn't raise this directly in the discussions either at WP:VPI or WP:Iritalk - one of the problems with U5 is how it incorporates part of the WP:User page guideline by reference; the new U7 text, unlike the versions presented before, does so even more directly, plus calls it policy instead of a guideline. That page is a lot easier to change out from under us, both procedurally and socially, than WP:CSD itself is, and it already leads both taggers and administrators to delete material exempted both in the current U5 and in WP:CSD#Non-criteria. See this recent deletion review for an illustrative example; the only thing at all unusual or uncommon about that case was that the page creator somehow managed to find their way to WP:DRV. Pointing at supplementary pages for further explanation is fine, but offloading the criterion to one to this extent really isn't. —Cryptic 13:40, 30 September 2025 (UTC)

  Will think on the rest of this, but for now I've corrected the policy/guideline error. -- Tamzin^{[cetacean needed]} (they|xe|🤷) 14:13, 30 September 2025 (UTC)

  It's not quite a supplementary page, it's a guideline. WP:DEL-REASON#13 already references guidelines by saying "policy" thinking "policies and guidelines" and links to the User pages guideline. I agree in spirit to your objection and it isn't appropriate that CSD should point to a guideline section which starts with "Unrelated content includes, but is not limited to:" Even though the policy proposal says "unambigously" someone could still claim that an item which does not explicitly match any of the bullets is unambiguously excessively unrelated content (might as well be, but that's not the point). The proposal could make that a little stronger (a note could say that the "but is not limited to:" part does not count for the purposes of CSD). Another thing that could be done is upgrading those guideline sections to policy sections ({{Policy|type=section}}) by running a (concurrent ?) RfC for that. —Alalch E. 14:40, 30 September 2025 (UTC)

  Sheesh, every guideline-related word in Roget's has a specific formal meaning already. I meant it in the ordinary English sense, as in "linking to additional pages for further info". —Cryptic 14:47, 30 September 2025 (UTC)