Hey, this talk page gets a lot of activity and proposals and questions about the nature of different speedy deletion criteria and when they should be applied but I have a new criteria to propose that perhaps hasn't been suggested yet.

I spend a lot of time with expiring drafts and also reviewing pages tagged for CSD speedy deletion and I think there should be a general criteria, to apply to both Draft and main space, for articles that are just blatantly fictitious. These are not hoaxes, where someone tries to pretend something is true when it isn't, it is kind of similiar to A11, for things that are made up or invented but what I keep seeing are what I view as "creative writing projects". They are stories new editors have written, sometimes fan fiction, sometimes fairy tales, sometimes ghost stories, it comes in many different forms. But it is all clearly fiction, not intended to be take for factual information. I see AFC reviewers offer opinions on submissions that are like this, saying things like "not enough reliable sources" but if a new editor makes up a story about a summer camp or a mermaid, there will never be a reliable source for that content. I've seen poems, song lyrics, all of this content is more suited to a personal blog than an encyclopedia. So, for Draft space, they sit for six months and are then deleted because, honestly, these editors never return to work on this content, it's just something they wanted to share or express one day. If it's in User space than CSD U5 can fit this content (see WP:UPNOT) but sometimes this material is in other namespaces. Maybe there could be a CSD D5 for webhost content in Draft space.

A different but related criteria could be called "stories about my life". So, new editors tell you about their best friend in school or how they went to this great concert when they were 18 or who their favorite actors are and why. These can be factual but they are clearly not about sharing knowledge and are not encyclopedic. I've seen content like this sometimes tagged as a G2 test edit but there are really no criteria to cover "personal writing projects". There is discussion on deletion pages about autobiographies and while we see a LOT of autobiographies in Draft space, that's not what these are. Both instances are more like essays or personal sharing but they are not hoaxes, they are not incoherent gibberish, not really vandalism or original research either, they are simply writings that are out-of-place in an encyclopedia. Aside from waiting out the clock for six months, it would be nice to have a critera that fit this content and an informative template explaining why this kind of content is not acceptable in this collaborative project. Thoughts? Liz ^{Read! Talk!} 20:44, 20 January 2023 (UTC)

I know what you mean. I deleted one of those the other day (Draft:“The Waffle House has found its new host” Lore) under G3 - it was tagged as such, but it takes a great deal of forced pragmatic incompetence to see that as a "hoax" rather than just bad creative writing. Incidentally when I was a new editor here that probably would have been tagged as G1.

If not a new speedy criteria (I think modelling D1 as similar to U5 is probably the best option) maybe a PROD type thing for those drafts would be another option? For things which are clearly not meant to be encyclopedia articles and never will be. – filelakeshoe (t / c) 🐱 20:54, 20 January 2023 (UTC)

I've seen a fair amount of these, and while G3 catches some of them, that is more intended for false information masquerading as legitimate articles (especially those created in bad faith). Deleting them is certainly reasonable, and keeping never-to-be articles as "AfC drafts" seems unnecessary at best, but I'm reluctant to delete when no CSD clearly applies. As such, I believe that such a new criterion generalizing A11/U5 might work – to cover original content clearly not intended to be an encyclopedia article and that would probably qualify for U5 in userspace: things made up one day, fanfiction and original stories, fantasy sports results, fantasy maps/transportation/proposals, personal essays unrelated to Wikipedia.

However, I don't know if there are several discrete categories that would make such a criterion objective and uncontestable, and I also am under the impression that not all admins interpret U5 and G11 (and even sometimes A7, which such a new criterion would resemble most) the same. Very careful wording is required. ^Complex/_Rational 22:26, 20 January 2023 (UTC)

It’s an old idea to push for A7 to be expanded to to cover draftspace, but it doesn’t make the justification. There is no harm in leaving weak claims lying in draft, and sometimes they are a draft for something with a claim.

The case for A11 is stronger, but, for compelling cases, like “stories about my life”, fit better to “unsourced BLP”. If it names a probably living person, it’s a BLP. No draft can be expanded on a BLP in the absence of a source, so it shouldn’t be a draft.

The answer is to expand BLPPROD to draftspace. Although not a proper CSD, PROD in unwatched places like draftspace is a pseudo CSD, and so the WP:NEWCSD tests are appropriate. BLPPROD meets these on all NEWCSD criteria but “frequent”. To provide evidence for “frequent”, please nominate all unsourced BLPs at MfD. If they are not frequent, this is still a good use of MfD.

I don’t think there are many A11 equivalents in draftspace that don’t meet any of WP:G1, WP:G3 or BLPPROD. SmokeyJoe (talk) 07:47, 21 January 2023 (UTC)

My only concern with an expansion of BLPPROD is that one can link, say, an Instagram page or TikTok video supporting a "story about their life", such that there technically is a source present, but the draft would be no more worthy of publication than without such a link. ^Complex/_Rational 14:56, 21 January 2023 (UTC)

User:ComplexRational, it's not clear whether you understand that BLPPROD is applied to an unsourced page and then requires a reliable source before it can be removed. An Instagram page or TikTok video is not a reliable source.

If the original page had the Instagram link or TikTok video, then the author is making some effort, and they deserve a conversation about WP:RS, and I OPPOSE speedy deletion of the draft under an expanded A11. SmokeyJoe (talk) 05:38, 22 January 2023 (UTC)

I'm all for discussing WP:RS with draft creators. I understand how BLPPROD is used on articles; are you merely suggesting to expand BLPPROD and offer a more detailed explanation, even for cases that nobody (besides perhaps the creator) would agree could be a plausible article? And while sometimes social media citations have their place, they would clearly not be sufficient for such drafts.

Per below, do you think a case such as "Jack has a crush on Jill. She is the most popular girl in school^{[citation to Jill's social media page]}." needs to go through BLPPROD as opposed to being speedied? I think it could also be helpful to survey recent deletions to gauge how frequently a case like this might occur. ^Complex/_Rational 15:12, 22 January 2023 (UTC)

If the citation to a social media page was present before, then according to the BLPPROD method, BLPPROD can’t be used. Use XfD. A newcomer already with the concept that claims need sources is worth the time to explain things.

In draftspace, BLPPROD will be a pseudo-CSD. “Speedy” doesn’t mean “instant” or even “fast”. I don’t think there is appetite for an instant speedy NEWCSD for draftspace.

To survey frequency, search the MfD archives for “unsourced BLP”. It’s not quite frequent enough. SmokeyJoe (talk) 12:14, 23 January 2023 (UTC)

There's a lot of schoolboyish stuff in draftspace that is only borderline G3 or G10 and should still be deleted as soon as possible. From "Jack has a crush on Jill" to "Joe Schmoe is in the eighth grade and jerks off to pictures of Taylor Swift". To delete under G10 works if we assume it is there in order to bully, but sometimes the content could be autobiographical and not intended as attack. I always delete this kind of things (and they should not go through MfD or anything else that increases awareness), but I am unsure what criteria to use. Something more clearly defined to catch these ("stories about non-notable people connected to the page author"? "possible violations of privacy"?) could be helpful. —Kusma (talk) 15:37, 21 January 2023 (UTC)

@Kusma: As another example, I recently deleted Draft:Paulinka as G11 (it was tagged as G3), which was essentially a love letter, but that felt like a stretch of the criterion. I imagine this is not too uncommon and needs a clearer criterion. ^Complex/_Rational 15:46, 21 January 2023 (UTC)

Schoolboyish stuff, borderline G3 or G10: "Jack has ..."; "Joe Schmoe is..."; autobiographical; .... This sounds VERY applicable to "unsourced BLP".

If you are not sure what criteria to use, you SHOULD NOT be speedy deleting. MfD should be used to establish a record and precedent for SNOW deletion criteria for new or expanded CSDs. You should not misuse criteria to avoid the NEWCSD process.

"stories about non-notable people connected to the page author"? "stories about ... people ..." squarely fits "unsourced BLP" in every case that I have seen that is suitable for a NEWCSD, for which BLPPROD will suffice.

Please send borderline cases to MfD. If there is a tad personal information that must be included in the MfD, the MfD can be blanked later. Pointing to a borderline G10 page from MfD does not induce a burst of pageviews beyond what is expected from the number of MfD regulars. SmokeyJoe (talk) 05:47, 22 January 2023 (UTC)

I too understand what you mean. I recently deleted Draft:Kaylin Beifong under G2, although I realise it's a stretch. That draft could probably have been deleted as a hoax, since the Beifongs are a fictional family in The Legend of Korra, and it's impossible that the user is a member of that family, but still G3 feels incorrect as well. Salvio ^giuliano 16:27, 21 January 2023 (UTC)

• In all of the above, nobody has explained why leaving those pages that are not BLP violations alone until G13 claims them is a problem? Why do they need deleting rather than ignoring? If there is an actual problem and an existing CSD criterion doesn't clearly apply then nominate it at MfD (remember that if it's unclear whether a page meets a speedy deletion criterion then it doesn't). Thryduulf (talk) 21:21, 23 January 2023 (UTC)

  For the same reason that the fake game show pages that were rampant for years are a problem. It's actively abusing Wikipedia for purposes entirely unrelated to building an encyclopedia, and allowing them to go through the standard G13 process wastes the time of both reviewers and people who check for salvageable drafts. In the case of the fake game show pages it was mostly happening in user space, so U5 basically resolved it. These sorts of pages are always deleted at MfD, they're not salvageable through normal editing, giving them wider attention is actively detrimental since it begets more creations along the same lines (as the fake game show problem demonstrated, a lot of them were time sensitive and going through MfD was about as effective a deterrent as watching a horse run from a barn before locking the door), and giving them a free 6 month pass gives the misguided impression that such pages are at least tolerated. That said, I'm also not sure how to capture that in a general criterion; if I could get the wording down I would, and if anyone has a starting point I'm happy to help refine it. The Blade of the Northern Lights (話して下さい) 23:04, 23 January 2023 (UTC)

  Make as many MfD nominations as you can using the proposed new CSD criterion. For U5, it was “NOTWEBHOST violation by a non-contributor”. G13 relied on the count of tens of thousands of abandoned drafts, too many to review, and that they included copyrights and BLP violations.

  The next CSD, a pseudo CSD, BLPPROD in draftspace, can be demonstrated by many identical MfD nominations “unsourced BLP”. It is appropriate to allow any interested editor one week to add a reliable source. “Unsourced BLP” nicely covers a multitude of sins, including borderline G10 and fake information associated with a person.

  I agree with Thryduulf that there what’s been no case made that there are ongoing draftspace problems in large numbers that are not BLP violations. SmokeyJoe (talk) 02:08, 24 January 2023 (UTC)

  So, unsource BLP requires that there are absolutely zero references and external links. But if an editor makes a page about a day in their life and includes references of the places they went to, that no longer qualifies as unsourced BLP. - UtherSRG (talk) 11:46, 24 January 2023 (UTC)

  Feel free to read WP:BLPPROD. All the same principles apply. I see no reason to alter any part of it.

  New editors attempting a sourced draft are quite a step above drive-by dumpers of BLP violation. SmokeyJoe (talk) 20:58, 24 January 2023 (UTC)

  If something has a 5% chance of being a BLP violation and a 0% chance of becoming an encyclopaedia article, immediate deletion is the way to go. Draft space is not a free webhost where you can post love letters, complaints about your math homework or random graffiti. —Kusma (talk) 12:15, 24 January 2023 (UTC)

  CSDs require a match to the CSD, not your probability analysis. If in doubt, use XfD. SmokeyJoe (talk) 20:59, 24 January 2023 (UTC)

  I use XfD when I have doubts about whether a page should be deleted. —Kusma (talk) 21:57, 24 January 2023 (UTC)

  CSD is not about whether you think a page should be deleted, it requires an exact match to at least one of the criteria. The criteria list the only circumstances in which the consensus of the community is that the page should be speedily deleted. Any other page should go to XfD. Thryduulf (talk) 22:31, 24 January 2023 (UTC)

  As several admins above have stated, there are instances of pages that should be deleted and should perhaps not go to XfD, yet there is opposition here to fixing the CSD to align policy with practice with arguments like that these pages do not appear often at XfD. WP:BLPDELETE allows many such deletions even if not covered directly by WP:G10 (and of course, there is always that other policy). —Kusma (talk) 23:02, 24 January 2023 (UTC)

  If something is too sensitive for XfD, go to WP:Oversight.

  BLPDELETE is an oddity that should be fixed. Maybe CSD#10a. It does not fit nicely within G10. BLPDELETE is a very poor example of speedy deletion authorisation from a non deletion policy page. It’s use under a best-fit to G10 sets this really poor precedent that speedy deletion by best fit is ok. SmokeyJoe (talk) 07:02, 25 January 2023 (UTC)

  As repeatedly explained on this page and elsewhere, IAR is never a suitable justification for speedy deletion because IAR is only for actions that uncontroversially improve the project, deleting something out of process (e.g. speedily deleting something that is not covered by a speedy deletion criteria) is always controversial and does not improve the project.

  Regarding things not going to XfD - if something is too problematic for XfD then request WP:OVERSIGHT. If something doesn't require oversighting and also doesn't meet G10 then its not problematic enough that XfD is a problem.

  WP:NEWCSD rightly requires that new criteria be both frequent and uncontestable, the way to demonstrate this is lots of nominations at XfD which always end in deletion. Thryduulf (talk) 10:28, 25 January 2023 (UTC)

  IAR says "If a rule prevents you from improving or maintaining Wikipedia, ignore it." It does not say anything about "uncontroversial". An objection against a specific IAR delete (which I don't actually perform very often, but I do see a need for them) must explain why that specific deletion was bad, not just say "all out-of-process deletions are controversial, hence bad". As you see from this discussion, there are many types of pages in draft space that are frequently speedily deleted by admins with a slightly bad conscience, and it would make sense to either align written policy with this practice or to criticise admins for specific incorrect deletions. —Kusma (talk) 10:47, 25 January 2023 (UTC)

  Kusma: you are saying that the deletion policies need to be rewritten to fit with out-of-process deletions that admins are currently making. Smokey and Thryduulf are saying that admins need to change their workflow so that their deletions fit with the deletion policies. — Martin (MSGJ · talk) 12:32, 25 January 2023 (UTC)

  Indeed as I wrote in WP:IARUNCOMMON, if you find yourself needing (or even wanting) to ignore a rule frequently either you or the rule is wrong and needs to change. I've seen nothing in the evidence presented that convinces me the rule is wrong. Thryduulf (talk) 01:19, 26 January 2023 (UTC)

I left a note about this discussion on WP:VPP#Discussion at WT:CSD, I figure some outside input could be helpful in fleshing this out. The Blade of the Northern Lights (話して下さい) 14:00, 26 January 2023 (UTC)

• This is yet another reason why draft space is an anti-wiki idea. Let's just do away with it and let deletion processes be used, where appropriate, on the articles created. Phil Bridger (talk) 19:14, 26 January 2023 (UTC)

Young students put all sorts of things in draft space naming real people, classmates and teachers. Other youngsters have written autobiographies complete with their addresses. I recently CSDed a draft that appeared to be a student writing silly things about a real teacher, nominated with G10 for lack of a specific G template, but it was deleted as G3. I did realize later I could have used {{db}} with a reason (disregarding the instructions that the reason has to be one of the existing ones). Years ago I used to nominate these as A7 with comments like "too much information about a minor child" or "children writing silly things about their classmates", but that no longer works in draft space. It would be very useful to have a G template for all the stuff that just does not belong on Wikipedia. StarryGrandma (talk) 22:57, 26 January 2023 (UTC)

• One place to start might be this content guideline, Wikipedia:Wikipedia is not for things made up one day.Andre🚐 23:42, 26 January 2023 (UTC)

I think everything User:StarryGrandma suggests that could remotely suit a NEWCSD, but does not meet G10 or Oversight, will fit better under BLPPROD-in-draftspace. SmokeyJoe (talk) 01:01, 27 January 2023 (UTC)

In CSD-Draft-version-of-A11/G3, anyone can tag the page and admins can go through, check if it conforms with policy to delete blatantly false or fictional material, and delete if so. Whereas BLPPROD if I understand it correctly will only apply to BLP situations and will let anyone delete old drafts after the waiting period. Andre🚐 01:18, 27 January 2023 (UTC)

BLPPROD only applies to BLPs, but quite broadly, not just to the subject of the page must be the BLP, any person tangentially connected can make it subject to WP:BLP.

It allows anyone to tag, with a sticky tag. The page author or anyone else has seven days to add one reliable source, or there’ll be a deletionist admin who quickly clears the expired BLPPROD category. The deletion criteria on an expired BLPPROD is so simple that the process is efficient. The applicability of G3 to a short draft can be very subjective.

A11, I support as extending to draftspace, but is it overkill? If it’s not BLP, why care? No reader is going to be unexpectedly misinformed by a draft page, as the namespace in the title, in huge font is kind of clear and meaningful.

I think one week is an appropriate time for a newcomer to be given to add a source to “Mrs Smith is a great English teacher”. SmokeyJoe (talk) 03:14, 27 January 2023 (UTC)

@StarryGrandma if you see pages (written by anybody) naming non-notable individuals, especially individuals who are or may be minors, contact Oversight (the information should not be easily accessible to even admins) rather than trying to bend speedy deletion criteria that don't apply. Ideally you should redact the problematic information (or blank the page if redaction wouldn't leave anything) using a bland edit summary. Thryduulf (talk) 01:58, 27 January 2023 (UTC)

It's encouraged to temporarily revdel or speedy-delete oversightable material while waiting for it to get oversighted: see Wikipedia:Revision deletion#Hiding oversightable material prior to Oversight. If it needs a speedy rather than a revdel, G10 or G12 may be an appropriate rationale. —David Eppstein (talk) 02:14, 27 January 2023 (UTC)

See Wikipedia:Village_pump_(policy)#Wikipedia_response_to_chatbot-generated_content for extraordinarily extensive background. "AI"s or LLMs (better term, since those "AI"s can be dumber than dirt), are being used to generate whole Wikipedia articles or portions of articles. When a whole article is produced, there are several issues: 1) the copyright status is unclear in many cases, though some LLM text generators do hold copyright. In any case, it is impossible to go back to the generator and compare to text there, as the output changes each time. 2) LLMs are great at making up stuff and making it sound plausible--it is highly likely that there are falsehoods in the output, and it would require the same level of effort as writing a new article to validate the text. And LLMs are great at producing fake references.

There is a policy being written for proposal soon. In the meantime, we're faced with articles that are somewhere between a G3 hoax and G12 copyright, produce significant verifiability issues, and in any event don't represent original work by the author of the article.

I'm proposing we have a new "G" variety for this issue. I know this is problematic until we have an actual policy, but I believe the above-listed issues are sufficient to have a wholly-LLM-generated article deleted even absent a new policy. — rsjaffe 🗣️ 23:45, 25 January 2023 (UTC)

Are there enough of these that a CSD criterion is warranted? Elli (talk | contribs) 01:28, 26 January 2023 (UTC)

(edit conflict) New speedy deletion criteria need to meet all of the four requirements listed at WP:NEWCSD, and while there might be a need for something in this area what we have currently isn't close to meeting three of them (uncontestable seems OK, once the others are fixed) -

• Objective - there needs to be some reliable and objective way of determining which pages are created by chatbots and which are not, and while I've not been following the discussions in detail it all feels very vague at present. Some of the detection methods seem to involve detailed comparisons with other sources which is wholly unsuitable for speedy deletion.
• Non-redundant - hoaxes and copyright violations are obviously already speedy deletable, are there any that are not copyright violations?
• Frequent - are there so many of these that normal processes such as AfD cannot cope? Especially when considering only those that are not copyvios and not hoaxes?

Thryduulf (talk) 01:29, 26 January 2023 (UTC)

In terms of quantity, yesterday I found 7 in about 10 minutes worth of looking. If we can by default claim copyvio then I have no problem using that criterion. There seems to be some uncertainty about the copyright status, however. As to objective, I’m only deleting those with a greater than 99% chance of being machine-generated using one of the detection websites. — rsjaffe 🗣️ 01:46, 26 January 2023 (UTC)

Wait, are you deleting them under some other criteria they're failing (since the ones I've seen have all failed multiple categories of deletion reasoning) or are you actually using this LLM claim as the deletion reason? Because I don't think there's any way to actually confirm the claim that any of these are LLM written. We can suspect, but if we get an article that actually meets our criteria and uses real sources, I don't consider "We suspect LLM text" to be a good reason at all. Silverseren^C 01:53, 26 January 2023 (UTC)

I’m using a tool by ChatGPT to evaluate the probability that the text is LLM-generated: https://openai-openai-detector.hf.space/. Using a cutoff of 99% probability makes the olds of false detection very low, and all the ones I’ve found so far were above 99.8% probability of being machine-generated. — rsjaffe 🗣️ 02:32, 26 January 2023 (UTC)

That...didn't answer my question. It really doesn't matter how many 9's you go into the decimals, we can't actually confirm it's LLM text, so our deletions should be based on other deletion requirements. Trying to make a subjective other tool to claim LLM written doesn't meet the CSD four criteria. Silverseren^C 02:35, 26 January 2023 (UTC)

That method is being accepted as an appropriate way to confirm it is LLM text. See Wikipedia:Large_language_models#Identification — rsjaffe 🗣️ 03:13, 26 January 2023 (UTC)

And if the editor says it's not LLM text and that they wrote it themselves, are we just going to say "No, you didn't" without actual evidence? Silverseren^C 03:20, 26 January 2023 (UTC)

And if it is LLM generated text, then it does not mean it is a copyright violation. Out of 2 that the OP tagged. One was a violation of a CV on a web site, and the other did not appear to copy anything preexisting. I am going to say no to such a speedy delete criterion. As LLM may be useful to us, and the problems are on the same scale as non-English native speakers writing, or topic-clueless people writing. Graeme Bartlett (talk) 04:47, 26 January 2023 (UTC)

I would generally be okay with a G-criterion to handle such cases if NEWCSD can be satisfied but the discussion is moot at this point because there is no policy in place that actually forbids such content. If and when the community decides that such content should not be submitted, then we can discuss how we handle violations. But we cannot establish a speedy criterion for content that is currently allowed (if it does not violate any other existing policies like copyvio). Regards SoWhy 08:48, 26 January 2023 (UTC)

I am working on improving WP:LLM so that it can be proposed as a guideline. Right now, what I have kind of pieced together is the idea of nominating them as G3 if they are trash -- I mean, utter trash. Not talking about if someone uses a language model to make a rough draft and fleshes it out here, more like the "I typed 'write a Wikipedia article about blungus' into the machine and blindly pasted the output into the edit box". Most of the examples from the AN/I thread have been deleted, but Draft:Gecko's was probably the best example: it had fake references and everything. I think that if you are going through a draft, AfC submission or new page and you discover that the references are made up, it should be some kind of speedy deletion. Whether this actually fits in G3 or if some kind of purpose-made criterion is necessary, who knows. jp×g 06:14, 26 January 2023 (UTC)

Sadly, G3 only covers blatant and obvious hoaxes. The kind of faked references found in Draft:Gecko's require work to determine that they are completely falsified; they are hoaxes, but not blatant and obvious ones. —David Eppstein (talk) 07:03, 26 January 2023 (UTC)

I agree with that per WP:DWHOAX but I think it's within the spirit of G3 if anyone can determine the fakeness of the sources easily (not saying that this was the case here). Regards SoWhy 08:44, 26 January 2023 (UTC)

What about a criterion for speedily deleting anything whose references are found to be fake? After all, that would be bad if the fake references were created manually, too. XOR'easter (talk) 18:44, 26 January 2023 (UTC)

That would work best as an extension of G3. —Kusma (talk) 19:38, 26 January 2023 (UTC)

This seems like a good path forward, we could add fake content and references to G3 under existing policies without waiting for LLM policy to be decided. –dlthewave ☎ 20:23, 26 January 2023 (UTC)

I nominated a couple as G3 and the admins reviewing declined to delete. — rsjaffe 🗣️ 16:48, 26 January 2023 (UTC)

I nominated one of those at MfD, Wikipedia:Miscellany for deletion/Draft:Iraqian economic influence in 2023. I agree that this kind of good-looking rubbish should have a G-speedy criterion, if we have a way to be reasonably certain that the page is LLM-generated of course. Fram (talk) 17:28, 26 January 2023 (UTC)

That is an example of a page that was created in article space and moved to Draft world. I wonder if this is an opportune moment to revise our procedures for draftification. For example, testing positive on one or more LLM detectors doesn't point to a specific source that was copied from, so a copyvio speedy delete would likely get turned down, as happened here. But it sure seems like good enough evidence to warrant immediate draftification, with an extra proviso that it can only be moved back to article space after a thorough check by someone other than the page creator. In other words, it calls for removing criterion 3 (there's no need to wait to see if constructive edits happen), and objections from the page creator should get no weight. XOR'easter (talk) 19:00, 26 January 2023 (UTC)

If we're worried that something is a copyvio we shouldn't be keeping it in draft space either. —David Eppstein (talk) 20:00, 26 January 2023 (UTC)

That makes sense to me, but there seems to be reluctance to drop the G12 bomb with only the results of LLM-output detectors to go on. XOR'easter (talk) 21:05, 26 January 2023 (UTC)

Very much support Rory's proposal. At the moment, we have no process for removing AI/LLM articles that are very obviously not suitable for an encyclopedia, but they fall through the cracks in the existing G and A CSDs: they're not patent nonsense, they're not blatant hoaxes, they're not unambiguous copyvios, they're often on subjects that appear notable, and they're not obviously invented. A CSD tag is likely to not be applied, or if applied to be turned down. With no process, these questions end up at AN, ANI or VP instead, waiting for someone to delete them out of process. This is a ludicrous state of affairs. With a forthcoming clean-up tag for drafts that might have something salvageable in them, we'll have a useful process for dealing with this bollocks and can go about our business free to pursue a life of religious fulfilment. — Trey Maturin™ 18:33, 26 January 2023 (UTC)

What objective criteria are you proposing to use to determine what content is AI/LLM-generated and which isn't? Thryduulf (talk) 19:02, 26 January 2023 (UTC)

There are an increasing number of LLM detectors. I would suggest figuring out which, if any, have consensus as reliable detectors and if at least one does setting a threshold (a very high one - I'd throw out 99%), and saying if it meets that percentage on a recognized detector it can be speedily deleted. Best, Barkeep49 (talk) 19:19, 26 January 2023 (UTC)

A field in a CSD-G15 template (and the clean-up tag for less obvious/potentially salvageable cases) to provide a link to our chosen detector and a percentage score would be an idea. It should even be possible to have a bot see the addition of the template, poll the chosen detector(s), retrieve a % figure, and add that and a backlink to the template for reviewing admins to, er, review. — Trey Maturin™ 19:58, 26 January 2023 (UTC)

Strongly support this as either a new CSD or an addition to G3. Draft:Gecko's archive was deleted under G3, however this was something of an IAR case because one could argue that it's a notable topic that could be a viable article if someone corrected all of the false facts, replaced all of the fabricated references and rewrote it in the appropriate style. If folks are worried about how to tell if it's actually AI generated, why not apply this to any and all articles that consist mainly of false/fake/fabricated content and references? –dlthewave ☎ 19:04, 26 January 2023 (UTC)

I am inclined to agree here: a criterion should probably look more like "article with completely fake references and gobbledygook text", and less like "created by this specific process". After all, these drafts would have been trash regardless of how they were made – it's just that there is a newly available technology which people are able to abuse in a way that produces this specific type of trash. jp×g 19:54, 26 January 2023 (UTC)

This is where the associated clean-up tag would come in, to catch AI articles that don’t qualify for CSD-G15 or have a glimmer of promise that a dedicated editor might be able to exploit. We certainly shouldn’t be speedying all LLM-derived articles, but we also shouldn’t be just leaving them be whilst we have debates about them in ever-more obscure places hoping that an admin will IAR delete them. A two-pronged approach, as we already have for all articles (speedy the obvious, tag the non-obvious), would cover most cases. — Trey Maturin™ 20:07, 26 January 2023 (UTC)

Thanks for making that maintenance tag. I like the idea of it, and I think it'd be easier to get consensus for than a new CSD. But I still think it's too long. I think we should shorten it to two sentences. I'd also encourage moving it out of userspace, to allow some bold editing and iteration by other folks. Hope this helps. –Novem Linguae (talk) 21:20, 26 January 2023 (UTC)

I went ahead and (with attribution) created Template:AI Generated which includes my attempt to get it to two sentences and to bring the wording more in-line with other content warnings we give (particularly about how to fix the problem). Courtesy ping @JPxG @Trey Maturin @Novem Linguae. Best, Barkeep49 (talk) 17:15, 27 January 2023 (UTC)

Okay, well, since we're on the subject, what should I tag this with? Draft:Draft:GPT-4 shows some obvious tells (i.e. is structured like a high school essay and lacks any real detail) and is scoring about a 6% on the analysis pages that have been linked earlier. jp×g 06:38, 27 January 2023 (UTC)

If you're sure it's AI-generated, I would BLAR it to the other draft at Draft:GPT-4. 6% doesn't sound like very good evidence though. Maybe just leave it for now. –Novem Linguae (talk) 06:53, 27 January 2023 (UTC)

Sorry: I meant that they returned 6% likely to be human-written (i.e. 94% likely to be GPT output). A rather important distinction I failed to make! For what it's worth, anyway (these analyzers are themselves rather seedy and seem to mostly be based on GPT-2). jp×g 09:35, 27 January 2023 (UTC)

I'm getting 99.98% GPT output for Draft:Draft:GPT-4. I'm going to try to CSD it. — rsjaffe 🗣️ 17:27, 27 January 2023 (UTC)

The difference between the two checks could be the COPYVIO material I revdelled. Best, Barkeep49 (talk) 17:35, 27 January 2023 (UTC)

And I've declined the speedy deletion because it doesn't match any current criterion valid in draftspace (it would likely be A10 in article space). Consensus for speedy deletion must come before speedy deletions. Thryduulf (talk) 17:53, 27 January 2023 (UTC)

Thanks for the info. So, until we get a changes in policy, if none of the current criteria match (bigger issue in draftspace than article space), tag it, and/or edit it to remove unreferenced questionable statements (WP:V). Right? — rsjaffe 🗣️ 18:34, 27 January 2023 (UTC)

If it's in draft space tag it and, unless there is an active problem, just ignore it. If there is an active problem take to MfD. In article space, take it AfD. Thryduulf (talk) 20:11, 27 January 2023 (UTC)

For the record, I have written an actual (crappy) article at GPT-4, so maybe this will solve the immediate problem of people trying to write extremely poor-quality versions of it with ChatGPT. jp×g 09:33, 27 January 2023 (UTC)

Currently, the opening paragraphs include the verbiage, " If an editor other than the creator removes a speedy deletion tag in good faith, it should be taken as a sign that the deletion is controversial and another deletion process should be used." I propose that this be modified to read, " If an editor other than the creator removes a speedy deletion tag in good faith, it should be taken as a sign that the deletion is controversial and another deletion process should be used. The one exception to this would be for G4, where non-admins cannot see the original deleted article. In those cases, the tag should remain until reviewed by an admin." This is due to the fact that the current verbiage says that another deletion process should be used. And at that point, since it has already been through AfD, Prod is no longer available, sending it back to AfD does not seem like a viable next step. If a third party removes the tag, and it cannot be re-added, no admin will ever have the chance to compare to see if there is a difference. There does not appear any inherent harm to the project for the tag to remain in place until admin-reviewed. Onel5969 ^{TT me} 22:20, 25 January 2023 (UTC)

Though the person adding the tag can't see the deleted article either. If we're going to change the wording like that for that reason, then logically wouldn't that just require that only admins can use G4 anyways? Since both the person adding and the person removing the tag have no idea what the article is like before and whether G4 applies. So both actions would be making a baseless claim. Silverseren^C 22:29, 25 January 2023 (UTC)

That means admins would have to do the new page patrol duties, as that's where these issues get detected. — rsjaffe 🗣️ 22:34, 25 January 2023 (UTC)

I think making placement admin-only would move in the wrong direction. Rather, we should add wording that clarifies that G4-placement is procedural and explain the process. Given that G4 isn't always a procedural shot in the dark (e.g. in cases where the nominating editor has full recollection of the previous version, or where there's a parallel version in draft space), it may make sense to create a G4.5 of sorts to allow for different text to be displayed depending on the context. _signed, Rosguill ^talk 22:40, 25 January 2023 (UTC)

Back before I became an admin I learned of several ways of viewing deleted articles to check against G5. The most reliable loophole has been closed but yes I agree that non-admin can assess against G4 some percentage of the time. What Onel's message caused me to think about was someone removing a G10 / G12 tag but decided that ultimately if that's what they were it could still be caught and speedily deleted when it got to an XfD and was seen by an admin. Best, Barkeep49 (talk) 22:43, 25 January 2023 (UTC)

The current policy is also confusing to the person who wrote the article/other editors who see the G4 speedy deletion tag. Especially for newer editors, they might remove the tag in good faith based on some memory of the pre-deletion article and be surprised by experience editors biting them saying that the tag is "procedural" and "waiting for an admin to review". As it currently stands, any non-admin editor can remove the G4 tag based on the immediate text of the template and the policy and the article may never get reviewed by an admin. EmeraldRange (talk/contribs) 23:42, 25 January 2023 (UTC)

G4 is a widely misused criterion, for some of the reasons noted above. We just had another instance come up at DRV. I'm not inclined to add more layers to it. If anything, I would rather see things added to the AfD queue with a "may meet G4" notation, and admins patrolling AfD enact those G4s if applicable. Jclemens (talk) 22:48, 25 January 2023 (UTC)

From my viewpoint this exception should probably be codified. If a non-admin removes a G4 template, this should not preclude requesting G4 again, as that should not be considered a situation when deletion is controversial and another deletion process should be used because a valid non-admin non-creator remover of the template can not, under most circumstances, be assumed to have good knowledge that the page is not a sufficiently identical copy, and has no solid reason to contradict the suspicion that the criterion applies, so there is no substantive controversy to be had – everyone would just need to wait for an admin to look at the deleted article and make the comparison. Sometimes, as Barkeep49 said, people will in fact have good knowledge, but there's no reason to generalize from that. Other proposals in the ANI discussion were to significantly soften the template, which seems go along with the idea, and a new software feature in PageTriage which would automatically gauge similarity. —Alalch E. 22:55, 25 January 2023 (UTC)

Though that does raise the requirement that admins actually check and compare the versions properly. Rather than in cases like what Jclemens brought up above. Should there be some sort of consequence for an admin who repeatedly deletes articles under G4 that were not substantially similar (and in many cases, blatantly so)? Silverseren^C 23:04, 25 January 2023 (UTC)

G4 is one of the most complicated criteria we have, and this means that pages are very often mis-tagged (I suspect due to people relying on the Twinkle dialogue instead of reading what the criterion actually says). Quite a lot of my G4 declinations could have been (and were) done as a non-administrator, such as due to the discussion having ended with a "soft delete" or "keep" or "redirect", the previous version having been speedily deleted or prodded, or that the discussion was quite clearly about a different person to the article being tagged. Looking through my G4 declinations, only two actually needed comparison with the previous deleted version (i.e. only two needed deletedtext), and one of them actually only kind of needed it, since it was clear that much had developed since the AfD. I therefore don't think this is necessary. Sdrqaz (talk) 23:33, 25 January 2023 (UTC)

If significant time has passed between delete and recreation it is also easy to see if new references or new claims of importance are added fro that time interval. So non-admins should be able to detag for that reason. If it is obvious that the AFD would no longer result in a delete because of something that changed in the meantime, there's no need to have another AFD. Graeme Bartlett (talk) 04:53, 26 January 2023 (UTC)

In addition to what others have said, if the previous page was deleted at XfD then the discussion should give an indication of the reason for deletion and in some cases it is very clear whether the reason for deletion no longer applies (e.g. if an article was unsourced and there are now sources; a redirect has a different target, etc). It would not benefit the project to prohibit non-admins from declining in these circumstances. Also, if someone other than the page creator removes a speedy deletion notice in good faith we should (in the absence of good reasons not to) treat it as controversial and require a discussion before deletion. Thryduulf (talk) 12:04, 26 January 2023 (UTC)

Additionally, bringing this from Template talk:Db-meta: I want to suggest adding language to the G4 template stating something like

If this article isn't a resubmission of previously deleted content, do not worry! An administrator will review your article.

Marking a page as reviewed and tagging it with a alarming notice that the article can be deleted at any time is non-intuitive to newer editors that this is a procedural tag which, in essence, asks for admin review. I was unsure if the tag meant anyone would swoop in and delete the article despite obvious (to me) reasons why the deletion no longer applies. EmeraldRange (talk/contribs) 12:59, 26 January 2023 (UTC)

This is wonderfully aspirational. Unfortunately, the fear you note above is neither strictly unwarranted nor irrational. Jclemens (talk) 04:17, 27 January 2023 (UTC)

• There have been several occasions where I, as a non-admin, have removed G4s because it is obvious even without seeing the deleted article that the current article is not substantially identical. This has usually been because the current article has content about things that happened after the previous deletion discussion. Phil Bridger (talk) 08:06, 27 January 2023 (UTC)
• Oppose There are many ways for a non-admin to determine that the deleted content is not substantially identical, e.g. Internet Archive, prior memory, new sources that did not exist at the time of deletion, etc. -- King of ♥ ♦ ♣ ♠ 00:55, 28 January 2023 (UTC)

R.E.: There is a discussion about changing—or making an exception to—#A10 criteria that is now taking place regarding reverted splits. Is this situation already covered here? I couldn't find anything. As I believe that this use is a reasonable interpretation of the existing #A10 rule, your input would be appreciated before I make any needed changes to the criteria section. ThanX, GenQuest ^"scribble" 16:27, 8 February 2023 (UTC)

I (and at least one other person) have left comments in the linked discussion rather than here to keep it one place. Thryduulf (talk) 22:35, 8 February 2023 (UTC)

I've removed links to and support for F10 in a whole bunch of places now, but from experience I will say that some are probably missing. I also started Wikipedia:Templates_for_discussion/Log/2023_February_12#F10 related templates about deleting the CSD templates as have been done for most previously removed criteria. I haven't touched twinkle, but removal has been requested by ‎Extraordinary Writ on the talk page. --Trialpears (talk) 07:50, 12 February 2023 (UTC)

Reason:

1. If non-free use rationale is disputed, it is not uncontestable since it is subjective what is "successfully addressed the concern".
2. This is a 7-day process so it is redundant as files can be PRODded.

GZWDer (talk) 13:49, 31 January 2023 (UTC)

• The mere fact something can be disputed doesn't make it subjective. In any case where a page is tagged for speedy deletion, under any criterion, you can dispute this and try to address the concern. If the reviewing admin still thinks the page qualifies for speedy deletion then it can be deleted anyway. F7 isn't redundant to PROD because PRODs can be restored on request for any reason or no reason at all, whereas an F7 deletion isn't likely to be restored unless you can show that the fair use rationale was valid. Hut 8.5 18:09, 31 January 2023 (UTC)
  • But that's not how Speedy deletion criteria are supposed to be applied. Any good faith editor can remove any speedy deletion criterion at any time, and that action should force a discussion. Why is F7 different? Jclemens (talk) 19:09, 11 February 2023 (UTC)
    • I assume the difference is that the uploader of the file can't remove an F7 tag (per For most speedy deletion criteria, the creator of a page may not remove the deletion tag from it) but can contest a PROD. No opinion on whether that difference is worth keeping a separate CSD around for. Extraordinary Writ (talk) 06:34, 12 February 2023 (UTC)

      Files deleted by prods can also automatically be restored on request. That should not ever be the case for copyvios. —David Eppstein (talk) 06:55, 12 February 2023 (UTC)

• I think this tag is meant for obvious cases like use stated in FUR is different from actual use (e.g. it says in the FUR that the image is used at the top of the article but the image appears at the bottom) and not for debatable cases like FUR states the file meets WP:NFCC#8 but in fact it does not, which are more suitable for FFD. The former problem often happens because someone clicks on buttons in an upload wizard without paying attention to which buttons are clicked on, or because an article is merged into a section of another article. --Stefan2 (talk) 07:28, 12 February 2023 (UTC)

  If it says in the FUR that the image is used at the top of the article but the image appears at the bottom then the problem can surely be rectified by just correcting the FUR? If there are examples of obviously invalid fair use rationales that are not currently speedy deletable but meet the WP:NEWCSD requirements then they should be proposed to be explicitly listed (similar to F7b), everything else should be prodded or go directly to FfD. While a file kept at FfD cannot be prodded, a restored prod can be taken to FfD and anyway no admin should be restoring a prod if the fair use rationale is obviously invalid. Thryduulf (talk) 11:04, 12 February 2023 (UTC)

  My gut feel is that it is rare that moving to the top can fix the problem. These cases are often something like the use of a non-free album cover for a movie soundtrack and is used in an infobox in the soundtrack section of the article about the film. Moving to the top would not be appropriate as it does not identify the film and usually there is already a poster there, and using for identification in a section never meets WP:NFCC#8. I also see this with corporate logos where old logos are replaced with new logos and the old logos are just shifted elsewhere in the article instead of being removed from the article. -- Whpq (talk) 13:56, 12 February 2023 (UTC)

  You've misunderstood what I was saying - I was suggesting changing the FUR to match where the image is used in the article, not changing the article to match the FUR. If you think an image does not add anything to the article you should remove it (subject to normal WP:BRD, or discussing it first if you think it will be controversial). If that image is fair use and it is removed from the only article it is used in then F5 applies. If you think the image does add something to the article then you should write a valid FUR for it. None of this offers any justification for the existence of F7d. Thryduulf (talk) 14:31, 12 February 2023 (UTC)

  If the FUR doesn't match the actual use, it's essentially the same situation as WP:F6, where a file is tagged for having no FUR at all. Sometimes it can be fixed by writing or editing a FUR, sometimes not. As long as we have a criterion for deleting files with no FUR, it makes absolutely sense to have a criterion for deleting files where the FUR doesn't match the actual usage. --Stefan2 (talk) 20:46, 12 February 2023 (UTC)

  I completely disagree that a FUR which editors dispute is or is not valid is the same as not having a FUR at all. The latter is an objective - the rational either exists or it doesn't - while the former is subjective and relies on differing interpretations of what is written. What I said above still stands - if there are examples of obviously invalid fair use rationales that are not currently speedy deletable but meet the WP:NEWCSD requirements then they should be proposed to be explicitly listed. The criteria as written would not be approved today and should be removed. Thryduulf (talk) 23:23, 12 February 2023 (UTC)

• After considering this, I suggest we Remove it: If it's obvious, G12 or F9 would apply. If it's not obvious, WP:FfD is thataway... Jclemens (talk) 15:01, 12 February 2023 (UTC)

  Looking at FfD, we could remove "What not to list here" III.5 and that would all that would be needed to align that page with this proposal. Or am I missing something? Jclemens (talk) 15:05, 12 February 2023 (UTC)

Every once in a while, I see pages like User:Yaseen Chaudhary show up on the U5 queue. Content's not very germane to Wikipedia but is quite short and the edit history contains a fair proportion of substantial edits (although I can't speak for their quality). Is this really within the remit of U5? Jo-Jo Eumerus (talk) 10:09, 12 February 2023 (UTC)

• That's very clearly "short autobiographical content" which is explicitly permitted by WP:UPYES so we don't even need to consider the nature of their contributions. That is nowhere close to being eligible for U5. Thryduulf (talk) 11:27, 12 February 2023 (UTC)
• If they had no project contributions, I would agree to it being U5, a drive-by self promotion dump, exactly the sort of thing U5 was intended for. However, they have contributions, and more bytes of them than this Userpage, so it is definitely not u5-eligible. “too short for U5”, your edit summary, is not correct. —SmokeyJoe (talk) 07:17, 13 February 2023 (UTC)

U5 is a very common criteria for newly created pages. While there are some pages that should certainly be deleted under it, like advertisements, fake articles (excluding drafts), and resources created for projects unrelated to Wikipedia, a lot of the pages that get deleted are just short biographies, many of which would be fine if the user was a contributor.

It's perfectly plausible that someone could create an account with the intent to contribute, and then create their user page before anything else. Immediately rushing to tag the page just because the user hasn't used the account to edit yet is quite WP:BITE-ey.

I think pages should only be tagged right away if the user is blatantly WP:NOTHERE. Otherwise, they should at least get a warning first. Zerbu ^Talk 13:41, 13 January 2023 (UTC)

​I wholly agree. A number of clear-cut NOTHERE user pages are often eligible for G3/G10/G11/G12, and I prefer to give benefit of the doubt unless there's a clear pattern to their contributions suggesting otherwise. Short biographies as a first edit are not harmful, and a number of established users probably made some of their first edits to their user page. However, I observe that not all users share these sentiments, and discretion varies among admins patrolling pages tagged for U5; this would run afoul of CSD being objective and uncontestable.

Moreover, per WP:UP#DELETE: A user's contributions that consist solely of a lone edit to their user page should not normally be speedy deleted unless it consists solely of spam or other speedy deletable material. Yet it appears that a number of pages tagged for U5 and subsequently deleted are exactly this.

I believe it may be a good idea to spell out what is and isn't eligible for deletion under U5, similar to how A7 is only appropriate for articles in a few narrowly defined categories, and accordingly explicitly codify in U5 what is written in WP:NOTWEBHOST and other egregious violations of WP:UPNOT as appropriate. For instance, this could include CVs, lengthy autobiographies, blog- or social media-like pages, "fantasy" content that would meet G3/A11 if in mainspace (such as fantasy football, original fanfiction), and pages intended for socializing or recreation unrelated to Wikipedia, along with (trivially) inappropriate pages that meet another CSD.

Also noting that this was (briefly) raised at WT:New pages patrol#Reminder (courtesy ping Liz), though no extensive discussion followed. ^Complex/_Rational 18:51, 13 January 2023 (UTC)

According to that rather unhelpful discussion, the people who patrol new user pages and tag them for U5 are not New Page patrollers. So who are they? —Kusma (talk) 19:06, 13 January 2023 (UTC)

Me, I tag a lot of pages with U5. There's probably at least a good 10 or so other users that I see regularly do so as well during my editing window (between 7am and 4pm EST). It really depends on which admin reviews a borderline U5 tag, some pages will be deleted and some won't. I try to always give users the benefit of the doubt, and I've started to err more towards not tagging pages that I know some admins would delete because I want to give people more of a chance on those borderline cases. As such, I think it could be beneficial to more thoroughly define what is and isn't eligible under U5 because there's different interpretations of what's currently written. It's not like users can review what was deleted from previous tags to understand situations when tagging is appropriate and when it's not. Hey man im josh (talk) 19:31, 13 January 2023 (UTC)

I've been thinking about this some more, and thought: perhaps, as an alternative to speedy deletion, such user pages could be blanked and replaced with {{userpage blanked}} (currently used mainly for drafts) or a similar template, which is less hostile than straight-up deletion. Zerbu ^Talk 23:00, 13 January 2023 (UTC)

I am inclined to think that U5 is the most overused speedy deletion criterium, and certainly the most subjective. In particular, I often see draft-like pages like User:Larsro/sandbox or short userpages being tagged as U5. Jo-Jo Eumerus (talk) 09:18, 11 February 2023 (UTC)

Perhaps it would be better to use the warning available in Twitter WP:TWINKLE (damn) that a user page is non compliant first. This is for other than egregious G11 or recreated content. -- Deepfriedokra (talk) 12:54, 14 February 2023 (UTC)

Yes, it's good to provide a warning, but what does Twitter have to do with this? Many Wikipedia editors, including myself, do not use Twitter, or any other sites that call themselves "social media", which is 180° opposed to the truth. Phil Bridger (talk) 19:01, 14 February 2023 (UTC)

I believe that User:Deepfriedokra was referring to Wikipedia:Twinkle. Donald Albury 19:34, 14 February 2023 (UTC)

Oh, OK, that makes sense. I have brain farts like that myself sometimes. I was trying to work out why an editor who usually makes perfect sense did not on that occasion and happened to hit on one of my bugbears. Phil Bridger (talk) 19:43, 14 February 2023 (UTC)

I read Deepfriedokra's edit as "Twinkle", and had to look hard at it after seeing your comment. Donald Albury 20:42, 14 February 2023 (UTC)

Maxim, casualdejekyll: moving the conversation about CSDP to its own section
Yikes. This coming Monday, it will have been a full year since P2 was last used. Besides those five deletions on January 16, it has been over two years since P2 was used.

P1 was used three times in 2022, not at all in 2021, and once in 2020. Let's take a look at those four deletions in the past three years. Portal:Ethiopia was deleted twice, once last August and once last October. The first time looks either like G4 (which does not need P1 to be applicable) or a misuse of CSD. The second time appears to have been a blank page, which would also be G2 as noted in the log entry (no idea why G13 was also listed). Portal:Content/Kenneth Nwanze was deleted under P1/G11, which again does not need P1 to be applicable. Finally, the one in 2020 appears to have been a blanked G11-eligible portal, which was also tagged with P1/A3 and P1/A7 for good measure. In sum, it appears that it has not been needed since before COVID times. Would someone with admin goggles be able to confirm?

For reference, A5 was recently depreciated for disuse when it was used 8 times in the past year, which is the exact number of times P1 and P2 combined have been used in the past year. I fully support deprecating both P1 and P2. MfD can handle an additional 8 cases over three years. I do not see the need to make CSDAs apply to pages in mainspace or portalspace, given that P1 was likely it was necessary zero times in the past three years. HouseBlaster^talk 03:47, 12 January 2023 (UTC)

Was looking at making this proposal as well earlier today. Here you have queries for all deletions under both criteria in portal space: P1 P2. Looking at all somewhat recent deletions I would either categorize them as either something you could have deleted anyway using a G criteria or a single sentence (examples include "Miami is a city in the United States.", "Someone please create a proper portal for this page. This is a country over a 100 million people, it needs a portal.") which I definitely can see why you would want to be able to unambiguously be able to delete, but given the frequency this is needed I feel it's reasonable to remove both criteria. --Trialpears (talk) 06:36, 12 January 2023 (UTC)

If we want to remove the portal criteria (and I support that given the lack of use) but still allow deletion of those examples, we could just amend A1 and A3 to say that these criteria also apply to portals. However given the frequency that is probably more hassle that its worth and sending them to MfD will not be any sort of burden. Thryduulf (talk) 09:52, 12 January 2023 (UTC)

I support deprecating both criteria. (Granted, any attempt to officially acknowledge that the entire Portal space is near-useless in its current state is unlikely to gain any sort of community consensus, so I'm just going to quietly grumble about it.) casualdejekyll 14:29, 12 January 2023 (UTC)

"Fixing Portals" is outside the scope of this discussion and page but I might take it to VP/Ideas at some point. casualdejekyll 14:48, 12 January 2023 (UTC)

If you do, be careful how you word things and avoid suggesting anything that could be interpreted as either mass (especially automatic) creation or mass deletion. I recommend familiarising yourself with at least a summary of the dispute if you aren't already - it really wasn't pretty. Thryduulf (talk) 15:29, 12 January 2023 (UTC)

• Question is G8 not applicable in some cases for portals on non-notable (etc.) subjects? For example if someone creates articles and a portal for their non-notable band and albums, and the articles are deleted as A7/A9, is the portal not dependent on the deleted articles? I honestly never understood the use case for P1. – filelakeshoe (t / c) 🐱 09:43, 13 January 2023 (UTC)

  I think the idea of P1 was that if people create their garage band/company advertisement article in portal space instead of article space, it should still be eligible for speedy deletion. Think Portal:Music/My garage band. (Back in the day, portals commonly had some original content not just derived from other articles). —Kusma (talk) 09:54, 13 January 2023 (UTC)

RfC: Should P1 and P2 be repealed as CSDs?

Apologies if this is something that's been previously discussed, but any thoughts on a CSD criteria for deleting orphaned file redirects created as the result of moving a file? —Locke Cole • t • c 05:36, 21 February 2023 (UTC)

Is this not covered by G6? Jclemens (talk) 06:01, 21 February 2023 (UTC)

You know, I thought that too, but then this happened. —Locke Cole • t • c 06:24, 21 February 2023 (UTC)

The policy here is WP:FILEREDIRECT, which says we should generally keep file redirects unless they're causing a problem of some sort. Outside of those situations, I'm not sure why we would want to delete file redirects—what harm are they doing? Extraordinary Writ (talk) 06:23, 21 February 2023 (UTC)

I dug through the history, killiondude mentioned that the devs apparently were strongly against deleting file redirects? Apparently there was some concern over attribution? I couldn't find anything more concrete (or that involved a dev), but I did only spend 15 minutes looking. —Locke Cole • t • c 06:37, 21 February 2023 (UTC)

Our files can be copied to places not on Wikipedia. They will have attribution pointing back to where the file comes from. So if the file was there for any long length of time, the chance is high it was copied and that the old name should be kept for attribution purposes. Also if the file was used in an article, then deleting the redirect stuffs up the view of the old revision of article. Also if some revisions of the file were deleted, they will be detached on a move. (sometimes though a history split needs to do this). Graeme Bartlett (talk) 10:03, 21 February 2023 (UTC)

@Locke Cole: That's how I remember it (see below), although that was of course 12 years ago. Still, I think the reasoning was basically what Graeme Bartlett outlines, i.e. that we should keep them for attribution purposes and that it would require subsequent edits (bot or manual) to change all uses on other pages on the project, thus creating a lot of additional work and edits with no tangible benefit. Unless the redirect is problematic, I would not supporting them and as Thryduulf correctly notes below, that is something that should be discussed at RFD for the few redirects where this might be the case. Regards SoWhy 17:16, 21 February 2023 (UTC)

It's definitely interesting. I think the concerns of attribution are likely irrelevant today, since deleted pages now show a log history which typically shows the target of the move as one of the entries. Regarding WP:FILEREDIRECT, it doesn't enjoy policy or guideline status, and should thus be no more influential than an essay. As Thryduulf notes below, this would likely fail WP:NEWCSD and I definitely agree it would fail #3 (frequent), but at least I have a better understanding of the "why". Redirects are WP:CHEAP, I just need to try to suppress the compulsion to clean up after myself. :P —Locke Cole • t • c 18:01, 21 February 2023 (UTC)

Even if the log shows the move target, deleting the redirect still needlessly breaks transclusions. This, combined with the fact that such redirects are seldom deletion-worthy, removes any reason to speedy delete them. Regards SoWhy 10:07, 22 February 2023 (UTC)

@Graeme Bartlett: if the file was used in an article, then deleting the redirect stuffs up the view of the old revision of article See, this was my argument for keeping old Templates, because once deleted, viewing old revisions doesn't work. But for whatever reason, we still nuke templates regularly that are (at least currently) "unused". As to attribution, even a deleted file still shows the deletion/move history, which points to the new file... is there something I'm missing here or not understanding? —Locke Cole • t • c 06:12, 22 February 2023 (UTC)

From what I've seen at RfD (and occasionally CAT:R3) this is a case where practice does not match policy; such redirects are frequently deleted. A year or two back I did a little bit of research on this, and found a proposal for the exact same thing here. SoWhy said: One of the MW developers threatened he will personally block anyone deleting such redirects. J947 † ^edits 10:35, 21 February 2023 (UTC)

I'm not sure what RfD's J947 is referring to as in my experience file redirects that don't meet R3 or F4 are only deleted if they are somehow actively harmful, with WP:FILEREDIRECT cited as the justification for keeping most others. How such a redirect is actively harmful usually needs to be explained individually. If you wish to propose a specific wording that meets all of the WP:NEWCSD requirements then go ahead, but I think meeting both points 2 and 3 is very unlikely. Thryduulf (talk) 12:10, 21 February 2023 (UTC)

I've just looked through the first page of RFD archives for discussions citing WP:FILEREDIRECT, of the 22 discussions (which occurred between mid 2019 and late 2022):

• 13 were closed as keep. 2 have subsequently been speedily deleted by Fastily under criterion F1, incorrectly as they were redirects not files. 1 was deleted per G8 when it's target was speedied per F5.
• 4 were closed as delete. 2 were determined to be actively misleading, 1 per the spirit of R3 and G7 and one that should have been no consensus (the nomination rationale was irrelevant and borderline incorrect, the other participants were evenly split about whether it was or was not harmful). None of these were uncontroversial.
• 3 were closed as no consensus, one of which I would have closed as keep as the only delete !vote was factually incorrect.
• 1 was closed as G8, another one was effectively a G8 but was closed as plain delete (the target was prodded). Thryduulf (talk) 12:10, 21 February 2023 (UTC)

Yeah the above is accurate. I was more referring to that from my impression of what file editors have said at RfDs, file redirects will often be suppressed or speedied. The deletion rate of a file redirect at RfD is much lesser than at CSD, which given all such file redirects are pretty much the same is a total failure of the system. J947 † ^edits 02:39, 22 February 2023 (UTC)

Previously: Wikipedia talk:Criteria for speedy deletion/Archive 83#Does G6 apply to redirects created via page move in the file namespace? * Pppery * _{it has begun...} 00:19, 22 February 2023 (UTC)

Thank you for that, genuinely. —Locke Cole • t • c 06:12, 22 February 2023 (UTC)

• WP:G9: In exceptional circumstances, the Wikimedia Foundation office reserves the right to speedy-delete a page. Deletions of this type must not be reversed without permission from the Foundation.

So I happened to be looking into some office action stuff and got curious about how often speedy deletion criteria G9 is used. From what I can tell no one has performed an office action deletion and called it a G9. If you look at this quarry you can see all times G9 is mentioned in a deletion summary. While there are 77 results none of them since 2007 are actually office actions but rather someone giving the wrong criteria or pressing the wrong button. I found some actual office deletions while doing this and they all just refer to it as an office action, usually linking to WP:OFFICE. This is probably because most wikis don't have a speedy deletion criteria for office actions. I checked Swedish, Spanish, French, Simple English, German, Portugese and Dutch Wikipedia and found that only French Wikipedia had such a criteria while Simple English removed their a decade ago. Smaller Wikipedias are probably even more likely not to have one.

I think it makes a lot of sense to do the same here, that is to remove G9 as a criteria. Even though the status quo doesn't cause any issues and this may very well be a solution looking for a problem I think there is value in limiting the number of unnecessary speedy deletion criteria. --Trialpears (talk) 05:31, 13 February 2023 (UTC)

Oppose. All speedy deletions should be authorised from here. If WMF log a link to WP:OFFICE, that’s good. If WMF do otherwise, they need education. SmokeyJoe (talk) 07:14, 13 February 2023 (UTC)

Oppose. There is an extremely strong community consensus that deletion without (the opportunity for) prior discussion (i.e. speedy deletion) is only allowed in a very limited set of circumstances, this page is a list of all those circumstances. Office actions are one circumstance in which speedy deletion is permitted by community consensus and so needs to be mentioned on this page. Your description of this as a solution looking for a problem is spot on. Thryduulf (talk) 14:16, 13 February 2023 (UTC)

Not mentioning a "G number" in the log summary doesn't preclude something from being deleted under that criteria. — xaosflux ^Talk 15:04, 13 February 2023 (UTC)

I agree with Xaosflux. A description that describes a CSD criteria without using the letter number code is still using that criteria. So an office action speedy deletion is using G4 just as much as "Attack page" in a deletion log is using G10. Best, Barkeep49 (talk) 15:08, 13 February 2023 (UTC)

An office action citing G4 is a flag for carelessness in the WMF Office. A link to WP:OFFICE for a WP:G9 is fine. These are synonyms. I have added a mention of WP:G9 at WP:OFFICE. SmokeyJoe (talk) 03:28, 14 February 2023 (UTC)

I think we should remove it and replace it with an entry in the list of other methods of deletion. To the extent that G9 is supposed to be an explanatory device for office actions it's useless because the WMF don't ever cite it when performing office actions, and WP:OFFICE gives a far better explanation of the process. Keeping it also implies that office actions are a type of speedy deletion and are subject to the other principles of speedy deletion, which isn't true. For example the WMF can delete pretty much whatever they like under WP:OFFICE, they don't have to stick to uncontroversial cases. Furthermore unlike all the other criteria ordinary admins aren't allowed to delete anything under G9 in any circumstances. Hut 8.5 08:54, 14 February 2023 (UTC)

If only G9 pointed to []Wikipedia:Office actions]], it would work just fine. SmokeyJoe (talk) 10:19, 14 February 2023 (UTC)

Perhaps the least invasive thing we need is G9 removed from any of the automated tools, so it cannot be selected by accident? Jclemens (talk) 21:20, 14 February 2023 (UTC)

Is G9 in automated tools in the first place? HouseBlaster^talk 03:23, 15 February 2023 (UTC)

I can verify that it's not in the standard version of Twinkle. I doubt that it's in regular automated tools due to the only being an office action kind-of-criteria. Clovermoss🍀 (talk) 05:44, 15 February 2023 (UTC)

It's also not in the dropdown list that appear when you click "delete" as an admin nor is it in the CSDHelper tool I use. Regards SoWhy 11:19, 16 February 2023 (UTC)

Well, then, my assumptions about where erroneous G9s were coming from appear completely unfounded. I'm a dinosaur who prefers Monobook, so most of the automated tools haven't worked for me this decade, even if I were inclined to use them. Jclemens (talk) 08:23, 21 February 2023 (UTC)

Actually looking at the list of deletions makes it clear what is going on. It contains:

1. 7 G8s related to Wikipedia:Articles for deletion/G9 (album) which aren't even claiming to use the criterion.
2. a batch deletion of 5 bot-created pages where an admin typoed "G9" instead of "G6" in a manual deletion summary
3. A false positive of File:Santa Clara County Route G9.svg, which was deleted per a FFD.
4. A deletion of Template:Db-g9 itself as vandalism, which resulted in a confused automated script filling in the criterion.
5. File:Andrew_Stewart_Jamieson.jpg was deleted because the article itself was deleted as an office action.
6. 3 instances of the (now-removed) feature where the content of a deleted article is autofilled in the deletion summary producing the words "G9"
7. Prior to March 2006, the number "G9" was used for what's now known as G6. Deletions under this use appear to have occasionally persisted until late 2007. * Pppery * _{it has begun...} 03:55, 20 February 2023 (UTC)

• Support I was originally planning to oppose this, but doing the above analysis reveals that the criterion seems to have never been used to describe an actual office action deletion, so listing it at Wikipedia:Deletion process similarly to the entry I added at Wikipedia:Deletion process#Copyright problems seems more logical than giving it a CSD code. * Pppery * _{it has begun...} 03:55, 20 February 2023 (UTC)
• Support If it's not used and not even referenced or linked to when OFFICE speedy deletions are rarely done, then there's no point on having it here. Because it just isn't used. It's a rule about nothing. Silverseren^C 03:59, 20 February 2023 (UTC)
• Oppose removal for now. Office deletions need to be mentioned somewhere in the deletion policies, and this is the traditional place. —Kusma (talk) 08:52, 21 February 2023 (UTC)
• Support. It is completely useless having this as a separate speedy deletion criteria. If this is supposed to be a policy it is useless, because editors here do not have the power to enforce policy on office actions, if it is supposed to be there for explanation purposes it is useless because no one ever quotes it and it is redundant to WP:OFFICE. The claim that we "need" this CSD to "authorise" office action deletions is complete nonsense, by design office actions are not subject to community review of authorisation - we could write whatever we want under G9 and the WMF would just ignore it. If really necessary add a line or two to WP:Deletion Policy explaining that the WMF may delete pages as an office action, but I doubt that's even necessary, we don't mention office blocks in the WP:Blocking policy for example. 163.1.15.238 (talk) 17:15, 23 February 2023 (UTC)
• Support - I was thinking of making this exact proposal myself as we seem to be on a roll deprecating speedy criteria that are never used at the moment. Completely agree with the IP above me. There is no more blatant a transgression of the "speedy criteria must be frequent" maxim than CSD G9, which is worded like some legalese backside-covering clause in a contract. The account User:WMFOffice has deleted a total of four pages in its entire history,[1] and none of the log entries reference CSD G9. All of them are the same kind of deletion of user pages of globally banned users which could be seen as uncontroversial housekeeping (though there may be disagreement on that). Whatever this is, it's not the same procedure described at WP:CSD where "administrators have broad consensus to bypass deletion discussion" as it does not concern administrators, and it's arguably not even part of enwiki deletion policy if it comes from the foundation, so it should for all those reasons be described elsewhere. – filelakeshoe (t / c) 🐱 22:53, 4 March 2023 (UTC)
• Neutral Genuinely torn here. It is ultra-rarely used. It plainly fails NEWCSD3. What gives me pause is that regardless of what happens here, G9 deletions will still take place. Repealing a CSD when it will still be used feels... weird. There is also the fact that enwiki (rightly or wrongly) does not like it when the WMF uses WP:CONEXEMPT, and I wonder if having an "official" CSD would help ease tensions. If WP:BANPOL had explicitly stated "the WMF may, at any time, for any reason (or no reason at all) decide to issue sitebans for any length", I highly doubt FRAMGATE would have been a dramafest (which might not be a good thing, given that it ultimately led to Fram being unbanned). TL;DR: I believe it has a very slight chance of having an effect of easing tensions and deletions will still occur regardless of what we decide here. OTOH, NEWCSD3. If this is adopted, I do believe we should notify the appropriate people at the WMF (whoever they might be) as a courtesy, even if they never officially used the shortcut in performing office actions. HouseBlaster^talk 04:59, 5 March 2023 (UTC)

This comes after a discussion at Wikipedia:Deletion review/Log/2023 February 27#List of mass shootings in Australia where I have advocated for the potential restoration of List of mass shootings in Australia which was created by a blocked user and thus speedily deleted. However, there seems to be consensus that the page topic is notable and it was decently sourced. However, my request to restore the page since it was unproblematic and potentially a useful article has been somewhat rejected by fact that many take G5 to mean that any page created by a sockpuppet should be deleted.

Echoing my discussion from the original argument, I do not see that interpretation from G5 at all. The literal text is that they are eligible for deletion, which is different from that they absolutely need to be. (Some speedy deletion candidates may be just a few edits away from being safe and that is the case here). I think that another policy, WP:BANREVERT, already makes it clear that good edits by sockpuppets are allowed to stay (although any editor that maintains those edits takes responsibility). So why would page creation, which is just another type of edit, be different? Obviously, an admin can check the deleted page to check for violations of neutrality, verifiability, or biographies of living persons. But outside of that, if an uninvolved editor sees use in the page, I think it should be made clear that the page can stay. It would be stupid to build it from the ground up when another version exists that was deleted solely due the creator. Otherwise, WP:BANREVERT needs to be amended as well to say that page creation is exempt from being allowed to stand (for whatever reason).

I can already pre-emptively see someone arguing about the second rule of CSD: It must be the case that almost all pages that could be deleted using the rule, should be deleted, according to consensus. However, even with my change, that still applies for G5. Most cases will result in deletion. This proposal refers to it being deleted and then undeleted. Why? I Ask (talk) 00:10, 2 March 2023 (UTC)

Agreed on the proposed amendment. On a personal note, something I've stated many times before, G5 remains the worst of all the CSD criteria and exists only to allow certain editors to make Wikipedia worse as some sort of revenge spree against someone for socking. We are here to make an encyclopedia, not to try and have vengeance against people that break the rules. If an article is notable, well-written, neutral, and doesn't have any other issues, such as copyvios or false/hoax information, then deleting it is very close to just being outright vandalism. Honestly, if everyone who socked only made such articles, that would be an amazing result. Silverseren^C 00:20, 2 March 2023 (UTC)

There may be many indefinitely blocked people who might create decent articles if they were to sneak in under new names and we allowed it. We don't mourn those because if they're indefinitely blocked, it's because net net their activities on Wikipedia were harmful. And blocked has to mean blocked, not "please sneak back on under another name if you plan to write a good article for us". It isn't "vengeance", it isn't as though sportsmanship is involved here. Deleting an article that a blocked user has created is no worse than the absence of other articles we might have if other blocked users were to pull the same stunt. If the person wants to return and create good articles and not cause net harm while doing so, let them follow the unblock process and make a decent and convincing request. The only exceptions I make are when I see the sockpuppet has fixed something that was broken. In those cases, I'm certainly not going to re-break the article. But even in those cases, when I review the sock account, there's typically been one or two real corrections and 20 acts of vandalism or erroneous edits. Largoplazo (talk) 03:14, 2 March 2023 (UTC)

Deleting an article that a blocked user has created is no worse than the absence of other articles we might have if other blocked users were to pull the same stunt.

This is an empty statement. You could extrapolate such a claim towards literally any article. Obviously removing an article on a notable subject that is properly written with no issues is a bad action. It doesn't matter the reason for it. If the article would meet all the requirements if a non-banned user had created it, then removing the article is harming Wikipedia. Full stop. Because, as I noted, we are here to build an encyclopedia. That is the purpose. And deleting every article or edit a banned editor made is not for the purpose of improving Wikipedia. It is for the purpose of "obey the rules, listen to our authority", which is not about making an encyclopedia, but demanding control. Blocking such users when they are discovered is fine and, if they indeed made any vandalism, that should be removed (as it would be if it was done by anyone and was vandalism). But removing positive content is not for the purpose of building this encyclopedia. Indeed, if a banned user could manage to make only positive contributions to Wikipedia and was never found out, that would be a net positive and a great thing. Indeed, statistically, there are many that have likely accomplished that to this day. Silverseren^C 03:25, 2 March 2023 (UTC)

Fine. Let's get rid of blocking. Everybody can do all the damage they want as long as they placate us by writing one good article once in a while, or at least hold out the promise of doing so. Fantastic.

It's almost as though you imagine that Wikipedia would have 50% more wonderful articles than it does now if only we let blocked users sneak good ones past us. It's the tiniest drop in a bucket, and for the greater overall benefit. Largoplazo (talk) 03:43, 2 March 2023 (UTC)

So you don't actually have a reason against this? No one is advocating that blocked users shouldn't continue to be blocked and banned from editing (even if they make good edits per WP:BMB). What we actually are saying is that good-faith editors can request the undeleting of a page when the page itself is fine. If the page is a nice B-class on a notable topic, but only edited by one person who is banned, why delete the article? To prevent them from returning? That doesn't seem likely to work (in fact, it does the opposite, from what I know; please prove that deleting their work demotivates rather than motivates them to try again). It is needlessly WP:POINTY, and it overtly does not help the encyclopedia on the content side of things. Why do you think that having the ability to go through the WP:REFUND process where an admin can look at the page for problems to potentially restore a fine page is bad? Why? I Ask (talk) 03:50, 2 March 2023 (UTC)

You should re-read what I said. I said we should keep blocking them whenever found. Particularly because they generally get found out when they go back to their old actions that got them banned, usually non-neutral edit pushing. But if a banned user can manage to avoid the topic area or whatever issues lead to the ban, contribute positively, and is never found out to be said banned user because of that, then it's a net positive because they've stopped the actions they were banned for and are actually trying to improve Wikipedia. If they can't manage that, then as I said, they'll be discovered and should be blocked once more. Silverseren^C 03:52, 2 March 2023 (UTC)

• I do not think that a single user asking for undeletion should be sufficient to restore the article to article-space, without any further qualifications. Adding this would allow any user who disagrees with G5 deletions to subvert the whole process by requesting all deleted articles to be undeleted en masse whenever a G5 deletion happens. If you're going to modify the criteria to add a backdoor making G5 useless, the correct thing to do is to build consensus to remove G5, not to trick other editors into agreeing to make it useless. —David Eppstein (talk) 00:22, 2 March 2023 (UTC)

  That is a ridiculous hyperbolic example. If an uninvolved editor wishes to support the content of the deleted article, then they are already de facto allowed to. Silverseren^C 00:26, 2 March 2023 (UTC)

  Sure, but writing it into the rules that they can always do so, with no safeguards, mean that they can always do so for all G5'd articles, and therefore that people who feel that "G5 remains the worst of all the CSD criteria and exists only to allow certain editors to make Wikipedia worse as some sort of revenge spree against someone for socking" could take revenge themselves about G5 by making such a request for all articles. —David Eppstein (talk) 00:33, 2 March 2023 (UTC)

  Then they would be blocked for disruptive editing, duh? Someone still has to review the requests at Wikipedia:Requests for undeletion. They can decide if someone is acting in bad faith. Just because there is no "safeguard" is not a compelling reason. I could tag every article for deletion if I wanted to in retaliation. Why? I Ask (talk) 00:38, 2 March 2023 (UTC)

  All you'd have to do is say that editors in good standing can make requests for undeletion of G5 deleted articles and take responsibility for the content, but should first request the undeletion from the original admin who deleted the page. Seems like simple enough wording to me. That is already the de facto process that exists and explaining it explicitly would actually be helpful, since many don't know about the contacting the original admin part. Silverseren^C 00:43, 2 March 2023 (UTC)

• Oppose as obvious forum shopping. * Pppery * _{it has begun...} 00:23, 2 March 2023 (UTC)

  How is it forum shopping? I am asking for there to be a change in the way G5 is applied and providing an example of how I think it should be applied. Your lack of good faith is terrible. Why? I Ask (talk) 00:30, 2 March 2023 (UTC)

  You've made a minority-of-one argument against at least 3 other editors that this article should be restored at the DRV, and are now trying to get policy changed to endorse that argument even though it's clear it doesn't have consensus (and edit warring on a policy page to remove similar language that resulted from a RfC). That looks like forum shopping to me. * Pppery * _{it has begun...} 00:37, 2 March 2023 (UTC)

  Because my argument is actually supported by a policy in WP:BANREVERT while the others are arguing over text that does not exist. I am requesting undeletion to keep the useful edits. I am taking responsibility for the article (per WP:PROXYING saying that: Editors who reinstate edits made by a banned or blocked editor take complete responsibility for the content). What I am doing is supposed to be allowed per those two policies. G5 just currently does not say it (but it also doesn't say the opposite), so I am trying to rectify it for clarity. Why? I Ask (talk) 00:40, 2 March 2023 (UTC)

  Just start the article yourself without using the deleted content. Simple enough, because if you're really checking the banned user's work with a fine toothed comb, it's no additional work for you. Jclemens (talk) 02:03, 2 March 2023 (UTC)

  Why would I do that when restoring it is a few clicks? This also doesn't account for future needs. Why? I Ask (talk) 02:19, 2 March 2023 (UTC)

  Because I'm assuming good faith that you're not so reckless to assume that banned user edits can ever just be undeleted and assumed to be good to go without thorough checking. If you don't understand that, I can't help. Jclemens (talk) 03:08, 2 March 2023 (UTC)

  When did I ever say they were going to totally be unchecked? The admin has to review it for disqualifying things (e.g., copyright and BLP violations). Otherwise, they are good to go. We have banned users that were once featured article makers. It's not that hard to say that their edits are helpful even if their presence is not. Why? I Ask (talk) 03:19, 2 March 2023 (UTC)

  Again, your naivete is not winning me over here: If it's complex, it needs as much work to vet as it does to recreate. If it's simple, it's trivial to recreate. The number of ways banned user contributions can be problematic appears to be outside your experience. Again, we're not talking underlying policy, just the wording of G5 to point out how to resurrect content. Jclemens (talk) 07:52, 2 March 2023 (UTC)

  I've got it! You don't have an actual consensus-based reason! This is a list of shootings. It's literally numbers and dates that I don't want (or need) to input again because it's time-consuming. It had already been done perfectly before. You also haven't even provided how restoring this page would be problematic. You've just maintained that it is because you don't seem to actually have a reason. And I'm sorry, but starting an article from scratch is way harder than reviewing and revising one that may have objectionable content. I kindly suggest reviewing Paul Graham's hierarchy of disagreement. Your rationales are rife with ad hominem and never go above a contradiction. Simply saying "no" because of "experience" is not an argument. Why? I Ask (talk) 08:06, 2 March 2023 (UTC)

  The fact that you have repeatedly demonstrated your ignorance of the why behind the ban policy does not mean that my noting this is ad hominem in any way. I'm not calling you ignorant, I'm calling your argument ignorant of the implications. You appear to have been here almost three years. The people you're disputing with, myself included, often have 5x that experience. Why do you think the consensus just isn't going your way? It could be we're all intransigent... or it could be we don't see that the change you're proposing is worth the potential damage it would cause. Jclemens (talk)

  If you think that it would cause more harm than good, then that's also fine. What is not fine (from you and others) is assserting that policy somehow supports your position. Your age on Wikipedia and opinions formed from that do not change what is currently written. It is true that an editor is allowed to request the restoration of a G5 as it currently stands. If you want to say that you do not want to add it here because of WP:BEANS, then I will not argue with you there. Why? I Ask (talk) 06:32, 3 March 2023 (UTC)

• What is the proposal actually? That the text of G5 should mention that anyone can ask for undeletion? That's unnecessary as it applies to the majority of speedy criteria. If you want such a page restored, you can just go to WP:REFUND. That's how things are at least in theory. In practice though, I acknowledge that G5 is one of the most haphazardly applied criteria (I've had difficulties both ways: admins refusing to delete creations of obviously disruptive editors, as well admins creating a mess by mass-deleting technical pages like dabs or redirects and then making the cleanup of that mess more difficult by refusing to restore the deleted pages).
  Or is it about this text that allows G5 to override prior AfD results? I agree that this should go, but it's unfortunately the result of a 2017 RfC. A lot of time has passed since then, so revisiting the issue in a new RfC may now be viable (incidentally, I couldn't help but notice among the supporters in that RfC the names of several now infamous indeffed editors). – Uanfala (talk) 01:39, 2 March 2023 (UTC)

  No, the text of the 2017 RfC is something else, but I do not want to start a whole other discussion while this is active.

  But apparently, WP:REFUND is vague stating that pages can be restored under certain speedy deletion criteria (and G5 is not one listed). Per several admins (such as User:Liz and User:HJ Mitchell) and at the deletion review above, there is a pervasive belief that pages deleted by G5 should not be restored (whether they are right is another thing, and I have shown policy-wise why I disagree). As you have also said that some admins have refused to restore pages, I think a direct statement saying that it is allowed (rather than being implied by text at WP:BAN) is probably in the best interest. Otherwise, we are just going to have this discussion somewhere down the line. Why? I Ask (talk) 01:49, 2 March 2023 (UTC)

  WP:REFUND doesn't list all the CSD's that can be restored there, it only lists the ones that can't, and G5 isn't there (though it used to be in the past). – Uanfala (talk) 01:54, 2 March 2023 (UTC)

  I see it lists the opposite of criteria that is not allowed to be requested and G5 is not there, so I question these admins' refusal to overturn the deletion. Would anyone oppose me adding clarifying text here, though? Why? I Ask (talk) 01:56, 2 March 2023 (UTC)

  Did you, in fact, ask for a WP:REFUND and get refused? All I see at the DRV is something else: asking for an opinion on the validity of the original deletion decision and getting an overwhelming consensus that it was performed appropriately. —David Eppstein (talk) 02:03, 2 March 2023 (UTC)

  @David Eppstein: The editor that opened up DRV already asked Liz, the person that deleted it, and she refused based on "rules" (what rules those were is unclear). And whether it's at WP:REFUND or WP:DRV matters little. All that matters is if you have policy-based reasons to overturn a deletion. Why? I Ask (talk) 02:06, 2 March 2023 (UTC)

  WP:REFUND admins have been quite explicit that they are hesitant to restore any G5'ed articles and prefer you discuss with the original admin first. Silverseren^C 02:12, 2 March 2023 (UTC)

  It isn't that easy. My request to refund Tekla Åberg was rejected at WP:REFUND, and then was only refunded by the deleting admin because there was a comment at the SPI supporting that action. — Preceding unsigned comment added by BilledMammal (talk • contribs)

• Oppose as not helpful in building the encyclopedia. As I alluded to in DRV, G5 is the way it is, and shouldn't be changed, on the basis of years of experience with banned users editing. Jclemens (talk) 02:01, 2 March 2023 (UTC)

  So you disagree with WP:BANREVERT and think that all banned editors should have their revisions removed? I'm sorry, but that is needlessly WP:POINTY and actually unhelpful when building an encyclopedia. I would prefer an actual policy based reason rather than you arguing that we need to enforce bans more. Yeah sure, we can do that, but removing a page that editors want and not restoring it after asked is not enforcing a ban. It's a waste of time. Why? I Ask (talk) 02:09, 2 March 2023 (UTC)

  Nope. I don't want the wording changed per WP:BEANS. Jclemens (talk) 03:09, 2 March 2023 (UTC)

  The way G5 is and has been for a long, long time is that you can indeed request undeletion and have the articles be restored if you are willing to take responsibility for the content. That is de facto how the system works and we should be explicit about how it is done. Silverseren^C 02:12, 2 March 2023 (UTC)

  I think it's a bit more nuanced than that. I have occasionally speedied articles under G5 and have restored them on request, but I have always taken account of both the article and the person making the request. When the article subject seems notable, the page is written constructively and the person making the request is an experienced user in good standing, then I have no objections to undeleting an article written by a sock puppet, but it's not exactly automatic. So, in short, I like the system as it is now: articles written by sock puppets may be speedily deleted, but administrators may also decide to restore them if it appears warranted under the circumstances, with DRV as the final safety valve/sanity check. Salvio ^giuliano 08:06, 2 March 2023 (UTC)

  I don't like giving administrators more power than they need. Obviously, as an admin, you reserve the right to refuse restoring anything that is copyrighted, libel, or anything that meets another CSD. But, you also don't get to single-handedly decide what is and isn't notable or even "warranted" (especially when there is another editor that objects). That is the purpose of community outlets like AfD and DRV. Thus if one good-faith uninvolved editor requests restoration it should happen. Take it to AfD if the page still sucks. That's what would have happened if it wasn't created by a sockpuppet. If there are worries about "gaming" the system, then by that notion, we shouldn't have AfDs either since sockpuppets have always disrupted there, too. Distrusting good processes because of bad actors is a terrible decision. And even as it stands with what you said, I still think there needs clarifying that people are allowed to request undeletion of G5. Several admins have thought otherwise as shown with this discussion.

  (Also, would you restore the page at the deletion review in question?) Why? I Ask (talk) 08:20, 2 March 2023 (UTC)

  Well, that's how I interpret WP:PROXYING: "[e]ditors in turn are not permitted to post or edit material at the direction of a banned or blocked editor (sometimes called proxy editing or proxying) unless they are able to show that the changes are either verifiable or productive and they have independent reasons for making such edits".
  G5 is a specific application of WP:BANNEDMEANSBANNED, whereby all edits made by blocked and banned users are supposed to be reverted. In these cases, since there is no "clean" version of the article, the page is deleted, so there is no need to have an AfD. Restoring an article on request is, on the other hand, an application of the exception described above and that is why administrators may restore an article, but it's by no means automatic. And you can't say that it's what would have happened if the page hadn't been created by a sock puppet, because it was and you can't overlook that. Salvio ^giuliano 08:44, 2 March 2023 (UTC)

  No, WP:BMB does not call for every edit to be overturned (as I have pointed out multiple times WP:BRV exists). It simply says that even if they try to make good edits, they are still banned since they have been held to be a net negative. It says absolutely nothing about needing to revert. And "independent reasons" is pretty vague, but as I have repeatedly asserted, as an "uninvolved" editor, they would be "independent" by default. And yes, it's pretty obvious with the existence of other pages by the same person (e.g., List of mass shootings in the United Kingdom) that they would be kept. But your entire reason is a gross misapplication of what WP:BMB actually says, so the second half of your argument about a "clean page" is untenable, anyway. (See why I don't think admins should have too much power.) Why? I Ask (talk) 08:54, 2 March 2023 (UTC)

  The fact that so many experienced users and administrators interpret policy differently should tell you that, at least, your interpretation isn't as clear-cut and obvious as you think,, especially considering policy is descriptive and not prescriptive anyway... And I agree that independent reasons is vague, and that's why I say that I take into account the article and the person requesting its undeletion, but I disagree with your opinion that it's enough that the request should be made by an uninvolved editor, because they are independent by default, since that interpretation of the clause basically makes the relevant condition meaningless. Salvio ^giuliano 09:25, 2 March 2023 (UTC)

  Sorry, but when the policy explicitly says the opposite of what you're saying, then it is your interpretation that is wrong. It's not like there's wiggle room like with some policies, it literally, directly says the opposite. Here's what you said: whereby all edits made by blocked and banned users are supposed to be reverted. See, but here's what actual the policy says: This does not mean that edits must be reverted. Sorry for being so pithy, but you cannot use the "interpretation" excuse here. And also, it doesn't make it meaningless, because that whole "independent" part is written to prevent meatpuppets and other bad actors from joining the fray. An uninvolved editor that has no ties to the banned user (after a check for suspiciousness) would be independent and then take responsibility for it. If there's confusion on what independent mean, we can start another discussion there. What do you think it means? Why? I Ask (talk) 09:38, 2 March 2023 (UTC)

  Nobody can be forced to revert edits by a banned user, but such reverts are allowed, and mass reverts are a standard practice. —Kusma (talk) 09:41, 2 March 2023 (UTC)

  Yes, such reverts are allowed. That is what I said. They clarified that they are supposed to which is an entirely different modifier. Why? I Ask (talk) 09:47, 2 March 2023 (UTC)

  If you want my opinion, the default should be to mass revert all edits by banned users without considering the content of the edit. (If a user is not banned, we triage their edits: revert the bad edits and keep the good ones. If a user is banned, we should not waste any editor time on triage). The policy as written makes this way of dealing with banned users fully within policy, but it also allows people to spend more time on banned editors than on non-banned editors if they think that is a productive use of their wikitime. —Kusma (talk) 09:56, 2 March 2023 (UTC)

  Yeah, and that's great, but it doesn't address my point. When another uninvolved editor requests the restoration of material reverted or deleted due to a banned user, they should get that. That does not conflict with what you just wrote. Why? I Ask (talk) 10:05, 2 March 2023 (UTC)

  Not to repeat myself, but they get it, as long as the administrator is satisfied that the editor in question is "able to show that the changes are either verifiable or productive and they have independent reasons for making such edits" (for an example, see here), but it's not and, as far as I'm concerned, should not be automatic. Salvio ^giuliano 10:09, 2 March 2023 (UTC)

  I feel like you're missing the key word of WP:PROXYING: at the direction of. That's the only time they actually need to be independent and verifiable as the policy is currently written. (I think I got that wrong earlier, too). If they aren't at the direction of the sock, then doesn't that mean they have the right to restore content? Why? I Ask (talk) 10:12, 2 March 2023 (UTC)

  You need to read all parts of the policy in conjunction, you cannot pick and choose the parts of the policy that support your interpretation. Bans apply to good and bad edits; good edits can be allowed to stand (which is why the policy says that it's not like they must be reverted), but to determine which ones can be allowed to stand you refer to wp:proxying, which includes the conditions I highlighted above. Salvio ^giuliano 09:42, 2 March 2023 (UTC)

• Oppose weakening G5. I don't mind occasional restorations of banned-user-created material to other people's userspace, but there should not be a codified right to do this. Remember that a large number of bans are done not because the banned user produces mostly problematic content, but because their interaction with others harms the community. We have to consider more than just the present state of the encyclopaedia, since we are more than that: we are a community that writes an encyclopaedia, and we need to have mechanisms to defend that community against abusers. G5 is one of those mechanisms. —Kusma (talk) 09:41, 2 March 2023 (UTC)

  If someone reverted a sock's edits (purely on the basis of sock) and another editor reverted it back because they liked those edits, those edits would stay, right? (Barring any sort of content dispute). The exact same thing applies with articles. You should have the right to revert back to an article you liked. It's just you need special permissions to do that. And the whole "defense" of the community is something I ain't buying. We defended the community by banning the perpetrator. Someone that sees the banned editor's content as encyclopedic shouldn't be refused a chance to restore that content. And thus, it should be noted that they have a right to request undeletion. Why? I Ask (talk) 10:16, 2 March 2023 (UTC)

  "Banning the perpetrator" doesn't do anything if it isn't enforced (so the community defense is done by making sure the ban is enforced). The enforcement for "you are not allowed to create articles" is to delete any articles that are created in violation of the ban. This content is not welcome here, no matter its quality. So you can't just ask for its undeletion, you would need to make substantial edits to the article so it is no longer made up only of edits that are violations of the ban. —Kusma (talk) 11:30, 2 March 2023 (UTC)

  There's nothing in the policy that says that. Sorry, but you would need to get larger community consensus for that idea that every edit by a blocked user should systemically be deleted. Why? I Ask (talk) 18:05, 2 March 2023 (UTC)

• Oppose there certainly isn't anything in the banning policy which says people have a right to restore contributions from banned editors, which is what this would establish. On the contrary, bans apply to good editing as well as bad, and if we've decided to ban someone then we've decided having them editing is more trouble than it's worth. Hut 8.5 18:19, 2 March 2023 (UTC)

  They aren't editing, so that policy you cited isn't applicable. Why? I Ask (talk) 18:46, 2 March 2023 (UTC)

  How is creating a page not editing? Hut 8.5 18:51, 2 March 2023 (UTC)

  Because this is talking about other editors that wish to use already made edits. The banned editor has already made their edits and had them deleted. There should be a place for a good-faith editor that still wishes to use those edits to do so. They are allowed to for undeleted edits, why not for pages that were deleted? Why? I Ask (talk) 19:00, 2 March 2023 (UTC)

  If any page is deleted for any reason then those edits are not available to be used anywhere. That argument is against deleting pages at all. Hut 8.5 19:16, 2 March 2023 (UTC)

  Yes, because there needs to be a history for attribution. I'm saying that good--faith users should be able to request undeletion (since in all other cases, they'd be able to restore reverted content). An undiscussed deletion shouldn't change that. Why? I Ask (talk) 19:19, 2 March 2023 (UTC)

  Yes, I get that you think that there should be some sort of right for editors to request that G5ed pages be restored, but I don't agree. We delete articles created by banned users for good reasons and they shouldn't be restored lightly. G5 isn't a mere piece of housekeeping like the other criteria which allow restoration on request. Hut 8.5 20:23, 2 March 2023 (UTC)

  You don't agree, and that is fine, but the policy is against you. At Wikipedia:Requests for undeletion, there is nothing prohibiting a user from requesting it. They intrinsically have that right. There are twelve criteria that are refused: F7, F9, F11, U5, A7, A9, A11, G3, G4, G10, G11 or G12. Notice the one that is missing? Because they are already allowed to do so! This is about the fact that editors have the right to request, not that it should automatically be restored. Why? I Ask (talk) 20:27, 2 March 2023 (UTC)

  You are rather creatively interpreting the instructions there. Yes, some criteria are excluded from consideration at REFUND, but it doesn't follow that pages deleted for any other reason will be restored. The important bit is the first sentence: Requests for undeletion is a process intended to assist users in restoring pages or files that were uncontroversially deleted. Anything which wasn't uncontroversially deleted is out of scope. And while I suppose you do have the right to ask, it will likely be declined. I've handled about 3,000 REFUND requests so please don't tell me I don't know how it works. I also don't see where you got the idea that the instructions at WP:REFUND are policy. They aren't. Hut 8.5 22:40, 2 March 2023 (UTC)

  I never said they had to be restored. I just said that they are allowed to request it, and that they should not be denied purely on the basis that it was deleted due to a G5 (since administrators each have different philosophies on this issue). And define "uncontroversially deleted". Basically, anything deleted without a discussion or legal issue (e.g., a PROD) is deemed "uncontroversial". Anything with a discussion is deemed "controversial". So G5 would apply as a deletion without discussion. Furthermore, it also says "certain speedy deletion criteria"; care to expound on what those are? Why? I Ask (talk) 22:50, 2 March 2023 (UTC)

• So, I notice that List of mass shootings in Australia is (as of my writing this) still a redlink. Is there a reason, Why? I Ask, you haven't just created the article and started writing it yourself? --Jayron32 19:03, 2 March 2023 (UTC)

  Why the heck would I waste time searching and adding sources and formatting a table when there was a perfectly good page to begin with. Do you have any reason not to restore it? The idea that we should just delete work and start over to get to the exact same result is baffling. The only reasons put forward is that "banned users should be reverted" (which is an opinion and opposed by policy) and that doing so "helps enforce bans" (which is another opinion not supported by any formal discussion, evidence, or policy). The gross misapplication of administrative power to override what users are both de facto and explicitly allowed to do is worrying. Why? I Ask (talk) 19:22, 2 March 2023 (UTC)

  As you have been told several times, you misunderstand this long-standing policy, which has been applied for as long as I can remember (and I became an admin in 2006). Your continued WP:BLUDGEONing assertions that everyone else is wrong are starting to be disruptive. —Kusma (talk) 19:30, 2 March 2023 (UTC)

  No, they are wrong. You have asserted something about a policy, but have not shown where it says that. If you think that a banned editor should have all revisions reverted, that is fine. It is allowed. What is not fine is not allowing a good-faith editor to independently vouch for those edits and keep them as the content was fine. The citing of WP:BMB does not say anything on this matter, and to say so is wrong. Why? I Ask (talk) 19:33, 2 March 2023 (UTC)

  I rest my case. —Kusma (talk) 19:40, 2 March 2023 (UTC)

  I rest mine as well. You still have not shown where editors are prohibited from requesting undeletion. Why? I Ask (talk) 19:44, 2 March 2023 (UTC)

  Why the heck would I waste time searching and adding sources and formatting a table... You don't have to do anything. This is a volunteer organization, and nobody told you to do any of that. You also don't have to have started this entire discussion, nor spent all the time you already have wasted contributing to it. I would have thought that time would have been better spent writing an article into Wikipedia that you were interested in. You don't have to do that. What I am saying is "The effort you spent on this discussion would have been better spent writing the article". If you don't want to write that article, then what the fuck is this all about anyways? --Jayron32 12:14, 3 March 2023 (UTC)

  It's about future cases, as well. This counterargument seems pretty awful. Firstly, I never said I had to, but since I want to, I shouldn't be denied to use material that was perfectly suitable. If I recreated it perfectly (copyright non-withstanding), then what was the purpose of deletion to begin with? Why? I Ask (talk) 12:22, 3 March 2023 (UTC)

• Oppose this change to G5 per WP:BMB, and per others' comments about gaming the system if anyone can request undeletion at any time. Personally I'm happy to provide copies of a G5-deleted article to any editor in good standing who asks, so that they can write their own article, but I won't restore banned users' work. Ivanvector (^Talk/_Edits) 19:27, 2 March 2023 (UTC)

  Can't someone already request undeletion? WP:REFUND does not prohibit a user from requesting the undeletion of a G5 edit; it is not one of the criteria listed. So it's already allowed, right? An admin is free to personally refuse, but what they are not allowed to do is say that requesting an undeletion is against the "rules". Why? I Ask (talk) 19:37, 2 March 2023 (UTC)

  WP:REFUND is only for cases where restoring the article would be uncontroversial, which is generally limited to housekeeping deletions. Restoring G5ed articles isn't usually considered uncontroversial. Certainly the reaction you got when you asked for this one to be restored shows it isn't uncontroversial. Hut 8.5 20:27, 2 March 2023 (UTC)

  You would first need to add G5 as one of the specified prohibited requests, then. Otherwise, see my above comment. It is allowed. Why? I Ask (talk) 20:29, 2 March 2023 (UTC)

• Comment: I think a more productive use of time would be to focus at what is allowed to be requested at WP:REFUND, otherwise there will never be consensus here. Why? I Ask (talk) 21:18, 2 March 2023 (UTC)

  There appears to be strong consensus here. It's just not in favor of your proposals. Please try to read the room when participating in discussions. And also avoid bludgeoning. All this tension is avoidable. –Novem Linguae (talk) 21:38, 2 March 2023 (UTC)

  The logic makes no sense: Currently, it is true that you allowed to request undeletion. So why is it not said that you can do so for clarity? Why? I Ask (talk) 21:44, 2 March 2023 (UTC)

  Indeed it's true. There's therefore no reason to adjust the policy? Especially for one single criterion? -- zzuuzz ^(talk) 22:35, 2 March 2023 (UTC)

  Because admins have repeatedly denied requests for restoration under that provision that it is against some sort of rule. And above, some editors have outright said that you can't request it. Therefore, there obviously is a current misinterpretation of what you are allowed to do, so spelling it out is beneficial. Why? I Ask (talk) 22:42, 2 March 2023 (UTC)

  I believe you've identified the flaw with DRV which is that it is mainly a test against deletion policy. It was a policy-compliant deletion, so it fails at DRV, regardless of the merits of the content. That's been that way for a long time. If you've come across other admins refusing to restore it, remember that admins restoring banned content also acquire 'complete responsibility' (or whatever the current wording is) for the content. I know of many banned users whose content I would never consider even looking at as I don't trust them. I still think you haven't presented the case for this highly singular addition. More words is more instruction creep. -- zzuuzz ^(talk) 23:18, 2 March 2023 (UTC)

  This isn't an issue with DRV, it's one with WP:REFUND. And several admins have already said there isn't any issue with this page other than the G5. If there are other issues, then it doesn't qualify. And admins are fine to refuse to restore. They can't say you're not allowed to request. (If you want my case for this singular page, ask me at my talk, please). Why? I Ask (talk) 23:54, 2 March 2023 (UTC)

• Comment. If an editor wishes to take responsibility for content provided by a blocked or banned editor they are explicitly permitted to do so. There is no policy that requires articles to be deleted just because a banned editor contributed to it, nor would doing so improve the encyclopaedia, so if someone wants to take on a G5-ed article then as long as they are acting in good faith and accept the responsibility for any errors in it then there is no reason to deny the request (assuming there are no copyvios, etc, but those should be deleted per G12 instead of/as well as G5). However, I don't think the CSD policy (and especially a singular criterion) is the place to note this, rather it's an issue with WP:REFUND. Thryduulf (talk) 10:25, 3 March 2023 (UTC)
• Oppose, many good points made above. One that hasn't been made is that, assuming the requester is a non-admin, they generally cannot assert that the deleted content is unproblematic, as they cannot see the article. All they have is the article name. If admins, who can see it, deny restoration, then there might be a good reason. A power imbalance for sure, but the more wily editors have a neat trick to get around this: they can use the time that might be spent sifting through a problematic article to see what is usable to instead write the article themselves under the same name! CMD (talk) 10:49, 3 March 2023 (UTC)

  Then the admin needs to clearly state: "it was not restored because..." Most editors can also get copies of deleted pages. But for using that content, per copyright policy, it needs to be undeleted to be usable. And the idea that creating an article from scratch is in anyway less time-consuming that simply checking over an already existing article and running it through a copyright violation checker is fallacious. Obviously, if some felt that was the case, they would not be requesting undeletion. Why? I Ask (talk) 10:59, 3 March 2023 (UTC)

  If your idea of checking for issues is to run the text through some particular copyright violation checker, then I can see why admins may hesitate to restore problematic content for your use. HJ Mitchell has kindly offered to provide all the relevant sources. If you don't actually want to write the article despite all the time spent on these talk pages, I would be willing to take a look at the sources myself. CMD (talk) 11:07, 3 March 2023 (UTC)

  No, that is not the only thing I'd do, obviously. And this is a list article. There is hardly anything to write. Literally, it's inputting dates and numbers into a table with a description. It is purely time-consuming. I also would like to ask you what problematic content there is on this page? Why? I Ask (talk) 11:16, 3 March 2023 (UTC)

  Happily, there isn't any problematic content on the page now, as it's a redlink. It is however a pretty obvious magnet for the sort of disruption that brings about discretionary sanctions, so I can see why it might have drawn the problems it seems to have drawn. If the sources HJ Mitchell provides are high-quality, then that's great! If they are not, I suppose there are probably high-quality sources to be found. CMD (talk) 11:28, 3 March 2023 (UTC)

  The only problem it had was that it was created by a banned user. Two other pages with the exact same scope exist, created by the exact same user (with one's content surviving AfD). Why? I Ask (talk) 11:35, 3 March 2023 (UTC)

  Unfortunately I can't confirm that, per my initial post here. However, why are there three pages with the same scope, and if there are why don't we redirect this one to them? CMD (talk) 11:57, 3 March 2023 (UTC)

  Same scope, but different countries: List of mass shootings in the United Kingdom and List of mass shootings in Switzerland. Why? I Ask (talk) 12:01, 3 March 2023 (UTC)

  Troubling, two articles lacking any sources establishing their core definitions, or sources speaking directly to the topic. I hope we are not seeking to create a third one. CMD (talk) 13:03, 3 March 2023 (UTC)

  As I understand it, G5 applies to page that were created by a banned user after they were banned. Pages created by a banned user before they were banned are not eligible for deletion under G5. The intent, as I understand it, is to deny recommendation to pages created by a banned user after they were banned. It comes down to timing. When were each of the three pages created? Donald Albury 15:35, 3 March 2023 (UTC)

• Oppose. Solution looking for a problem. Stifle (talk) 10:50, 17 March 2023 (UTC)
• I will say that uninvolved editors can always ask. If they are involved it will invite a block. But then the issue is: should admins ignore or action or deny the request. Graeme Bartlett (talk) 11:37, 17 March 2023 (UTC)

Strictly according to WP:G13, the deletion criterion applies to pages that have not been edited by a human in six months. This does not explicitly exclude sockpuppets of blocked or banned humans doing so in contravention of their block/ban. What is the official position here? If the intention is that such edits should not reset the clock, should this be made explicitly clear?

Apologies if this has been asked - I could not find this specific issue in the archive.

Dorsetonian (talk) 10:17, 19 March 2023 (UTC)

Well sockpuppets are humans, so yes this resets the clock. And I don't think this situation is nearly frequent enough to justify adding a rule specifically excluding them. Hut 8.5 10:38, 19 March 2023 (UTC)

The edits do reset the clock. If this causes issues, MfD and mention there that the page would have been a G13 if not for the sock. This should not be made explicit in the rules per WP:BEANS. —Kusma (talk) 11:13, 19 March 2023 (UTC)

Why are you responding to sockpuppet edits? They are probably doing it to play with you. Note your observation at the SPI case page, and per WP:DENY, do not start MfD discussions or WT:CSD. SmokeyJoe (talk) 11:18, 19 March 2023 (UTC)

No, but not worth correcting - they're draft space articles, if it takes a few extra months for them to be deleted it doesn't matter, and isn't worth the effort of good faith editors to identify and correct. BilledMammal (talk) 11:24, 19 March 2023 (UTC)

Agreed that, as currently worded, any human edits do reset the G13 clock, including those by sockpuppets. Counter BilledMammal I would say this is worth correcting, since it's an idiosyncratic exception to the general way that Wikipedia allows people to pretend sockpuppets never existed. * Pppery * _{it has begun...} 14:53, 19 March 2023 (UTC)

Where is the prior discussion that led up to this RfC? --Redrose64 🌹 (talk) 21:50, 19 March 2023 (UTC)

Good point. There was no need for the rfc tag and I will remove it again. Apologies for the drama. Dorsetonian (talk) 18:45, 29 March 2023 (UTC)

They would reset the clock. G13 deletes are supposed to be speedy and should not involve a sockpuppet investigation of contributors. So it is too much trouble to add that rule and implement it. If a sock does edit the page, then consideration has to be given to the whole lot being a sock creation, or banned editor creation. In which case a G5 can be used. If not a revert and additional wait is not serious. Graeme Bartlett (talk) 11:04, 20 March 2023 (UTC)

Yes, and this is not a problem per Billed Mammal and the first part of Graeme Bartlett's comment. As with everything in draft space, if the page is actively causing a problem then fix it, take it to MfD or alert WP:OVERSIGHT as appropriate. In all other cases ignore it. Thryduulf (talk) 15:26, 20 March 2023 (UTC)

Of course getting rid of WP:G13 altogether would mean that this is no longer a problem. It would also have the effect that draftification would no longer be a backdoor route to deletion, as it currently is. Phil Bridger (talk) 18:51, 20 March 2023 (UTC)

That's just crazy talk. The WMF is clearly going bankrupt and cannot afford the disk space for what's in draft space as it is! /sarcasm. Jclemens (talk) 05:21, 22 March 2023 (UTC)

U2 "applies to user pages, user subpages, and user talk pages of users that do not exist". The criteria instructs us to check Special:ListUsers. Special:ListUsers only lists people who have registered or edited locally. However, the {{db-nouser}} template links to Special:CentralAuth which provides global account information.

I'm looking at User talk:Aa, which is tagged as U2. They do not have a local account, but do have edits on the Kyrgyz language Wikipedia, and an account on the Japanese Wikipedia (no edits there). I am totally lost as to if this qualifies for U2 or not. ~ ONUnicorn^{(Talk|Contribs)}problem solving 20:12, 28 March 2023 (UTC)

• This qualifies as G5, which I'll be doing shortly, thus spoiling the question. Interesting question though. I'd say unless it relates to relevant activity somehow (which excludes an out of context welcome message), then it qualifies as U2. -- zzuuzz ^(talk) 20:28, 28 March 2023 (UTC)
• This criterion dates from before SUL was a thing so going back to the original discussions wont help. Given that accounts are created when you read pages on a wiki (tested by going to special:centralauth/Thryduulf and looking for a project I didn't have an account on (I chose the Walloon Wiktionary), visiting it, then looking at my centralauth list again. I do now have an account on that project even though I have never done anything other than read. This means that someone without an account here is definitely not engaged with the project, and so I think U2 should apply, regardless of whether the account exists on another Wiki. Off-topic for this discussion, but if you see someone leaving welcome messages for accounts that don't exist, they might be gaming auto (or extended) confirmed status. Thryduulf (talk) 20:36, 28 March 2023 (UTC)

  Sometimes an account is spontaneously created without visiting that Wiki. For example, Special:CentralAuth/Redrose64 lists 234 different wikis; of these, 160 are Wikipedia in various languages, and I have visited several of them, but I don't think all 160. Of the other 74, there are the expected all-language wikis such as commons, wikidata, meta, species plus varying numbers of language-specific wikis - wiktionary, wikisource, wikibooks etc. Some of these language-specific non-English wikis were linked as long ago as 2009, but I am absolutely certain that I have never visited any of the three non-English wikiversity wikis, nor six of the eight wikibooks wikis. Some are mysterious: until I analysed this list, I had never even imagined that bd.wikimedia.org, ru.wikimedia.org and ua.wikimedia.org might exist. What on earth are they anyway? --Redrose64 🌹 (talk) 22:03, 28 March 2023 (UTC)

  Wikis associated with Wikimedia chapters, for the record. The usual other cause of an account being auto-created is if an edit they made is imported, which is what happened to you, at least with the chapter wikis. * Pppery * _{it has begun...} 22:45, 28 March 2023 (UTC)

  This shows that someone having an account here does not demonstrate that they are engaged with (or even aware of) the project in any way, but it does offer further evidence for the converse - i.e. not having an account at en.wp demonstrates that they are not here. Thryduulf (talk) 13:37, 29 March 2023 (UTC)

• I've deleted ~250 pages under U2 in a year with ~85 nominations before that, and although I check CentralAuth before deleting, I carry them out if a local account does not exist (while keeping in mind the other exceptions, of course) – on a fundamental level, the pages in question are pages on the English Wikipedia, so the existence of a local account is what matters. If my understanding is out of step with the wider community's, I will amend my practices (and try to trawl through my deletions and restore where necessary). Sdrqaz (talk) 04:45, 31 March 2023 (UTC)
• It sounds like we're all in agreement that only local accounts matter, and as everything except {{db-nouser}} links only to local users the simplest thing to do is just change that template. Thryduulf (talk) 11:29, 1 April 2023 (UTC)

  I've changed the template. * Pppery * _{it has begun...} 15:05, 1 April 2023 (UTC)

A discussion at AN/I about articles being created by the various 'AI' programs being offered has shown that we're at the start of a problem.

At the time of writing, there are 84 drafts (mostly because they've been moved to draftspace) that other editors have tagged with {{ai-generated}}. There are a hell of a lot more, including in mainspace, that are untagged.

It has also shown that the editors adding these articles are unaware (or claim to be unaware despite the widespread news coverage and disclaimers on the chatbots themselves) that such articles are plausible bollocks.

Note that phrase: plausible nonsense. They're not blatant nonsense and they're not obvious hoaxes. But they are both nonsense and hoaxes.

At the moment, we're relying on admins to misuse CSD-G3 and delete them by waving WP:IAR. I contend that this is asking our admins to trust that a troll or even a well-meaning fellow editor won't drag them to various forms of public stocks (AN, ANI, ArbCom) for violating the letter, if not the spirit, of G3. This is having a chilling effect on admin action, and also causing even clear-cut articles to be taken to AN/I because we lack another way of dealing with this.

Yes, XfD exists, but this problem is exponential. If we don't act to remove these plausible nonsense pages quickly, XfD could easily and quickly fill up with dozens and then hundreds of these things, with plausible-hoax-nonsense-bollocks hanging about in mainspace (and being picked up and reused by AI content scrapers) for a week at a time. I'd rather that they were nuked immediately and WP:REFUND used for any human error, just like with every other CSD.

Therefore I propose

CSD-G15: "Page created using an AI large language model chatbot which has no use to an encyclopedia."

(Wording subject to community revision, of course). — Trey Maturin™ 16:49, 1 April 2023 (UTC)

Please checkout some of the recent discussions on the matter, if you haven't, mainly this one. The consensus seems to be no because we can't reliably determine that pages are created using LLMs. --Trialpears (talk) 16:54, 1 April 2023 (UTC)

I'm aware of that discussion – I commented in it a couple of times. This needs further discussion rather than being handwaved away with "oh, we talked about this a while ago". It's getting urgent now. — Trey Maturin™ 17:00, 1 April 2023 (UTC)

How does G3 not apply exactly? The pages are using fake references and thus clearly fall under hoaxes, as with many, many past hoax articles. The issue is not the LLM writing, the issue is the false information included that the editor who made the page did not correct. Silverseren^C 17:00, 1 April 2023 (UTC)

The admins who have been deleting using G3 have all said that they're aware that G3 doesn't quite cover it. There are also admins who are saying they won't delete because G3 doesn't quite cover it.

That could be an argument for increasing the scope of G3, of course, but I can just imagine the horrified reaction here from even suggesting such "scope creep". — Trey Maturin™ 17:03, 1 April 2023 (UTC)

Then those admins need to explain why G3 doesn't cover cases of hoax articles. Since this has nothing to do with LLMs, it has to do with articles having fake information and fake references, the definition of a hoax article. All I can say is that admins saying that are just wrong. Silverseren^C 17:06, 1 April 2023 (UTC)

The problem here is that these are not blatant hoaxes. These articles are plausible hoaxes. G3 simply doesn't allow for that, although I welcome your willingness to expand G3 to include that. At the moment, such unintentional plausible hoaxes have to go through XfD. Within a few months – if not a few weeks – XfD could, nay will, be bursting at the seams with these things. — Trey Maturin™ 17:16, 1 April 2023 (UTC)

If they are not blatant hoaxes then they should not be speedily deleted full stop, because a single admin cannot reliably determine whether it is or is not a hoax. This also brings up the related issue with LLMs - to date nobody has demonstrated a way that a single admin patrolling the speedy deletion category could determine, reliably, whether any given content was written by a human or LLM (and not all content that was written by an LLM should be speedily deleted. This means that your suggested criterion would fail three of the four WP:NEWCSD requirements.

Claims that XfD will be bursting at the seams with LLM-generated (or other) content have been made before and have never come to pass yet so forgive me for being sceptical of yet more hyperbole. Thryduulf (talk) 17:22, 1 April 2023 (UTC)

Well, we'll just have to see then. I will gleefully and childishly point you to this thread repeatedly and without apology when it comes to pass, because I am a very petty person, alas. — Trey Maturin™ 17:26, 1 April 2023 (UTC)

As noted in the previous discussion, being created by a LLM is irrelevant. If there is a problem with the writing that cannot be fixed then deal with it in exactly the same way you would if it was written by a human. For example, G1, G2, G3, G11, G12, A1, A3, A7, A9, A10, A11, AfD and Prod. Thryduulf (talk) 17:17, 1 April 2023 (UTC)

Which doesn't allow for the (clearly forthcoming) battles over copyright and plagiarism related to both the method of synthesising LLM content by "training" it on others' work and to the posting of it unattributed to our encyclopaedia. — Trey Maturin™ 17:24, 1 April 2023 (UTC)

Copyright violations can and should be speedily deleted under criterion G12, regardless of why it is a copyright violation. Thryduulf (talk) 01:14, 2 April 2023 (UTC)

Copyright violations in the form of copied text with all the words changed to synonyms can be difficult to detect, but the fact that we often have the source of the text listed as a reference makes it easier. Copied text, closely paraphrased from unidentified sources, as LLMs can do, makes it much more difficult. G12 is unusable unless we know what it was copied from. Preemptive bans on LLM text are easier. —David Eppstein (talk) 01:58, 2 April 2023 (UTC)

Still seems like LLMs have nothing to do with it and our regular rules cover the situation. If the text is unreferenced, then that's already a reason to remove it. And if the reference is fake or otherwise doesn't contain the article content it is being used for, then that's another reason to remove the text. Silverseren^C 02:05, 2 April 2023 (UTC)

Are you trying to suggest that we should tolerate Wikipedia editors who perpetrate close-paraphrased plagiarism, as long as their references check out and they are careful to give different references than the ones they copied from? —David Eppstein (talk) 02:09, 2 April 2023 (UTC)

No. If text is plagiarism then remove it, the nature of the plagiarism and whether it was plagiarised by a human or LLM are both irrelevant. If you aren't sure whether it is plagiarism or not then speedy deletion is the wrong tool, regardless of who or what wrote the text in question. Thryduulf (talk) 02:11, 2 April 2023 (UTC)

• Support this is exactly what speedy deletion is for. These AI engines are not capable of writing content suitable for Wikipedia. They frequently make up information or misunderstand things, but they are also capable of producing very convincing-looking content. The mere fact that some content was written by one of these engines means that we shouldn't have to evaluate it further. Hut 8.5 10:53, 2 April 2023 (UTC)

Oppose until we have formulated a way to determine that an article was written by an AI. Limiting it to only articles that explicitly say they were made by a LLM might narrow the criterium too much, but not having a qualifier could lead to deletions when an article only maybe was written by an AI. I'd also consider making such a criterion temporary; I would not categorically assume that AI will remain incompetent at Wikipedia writing forever. And if by the time the criterion expires the AIs are still not ready, we can prolong it then. Jo-Jo Eumerus (talk) 16:17, 2 April 2023 (UTC)

Oppose Definitely a growing problem, and likely to require some sort of specialised approach in the future, but this is still somewhat premature. Keep a well written draft of this criteria up your (metaphorical) sleeve though, it is probably going to be needed, but now isn't the time. Prerequisites to a successful new CSD criteria here would be:

1. A tool which uses a (potentially yet-to-be-developed) API or something to quickly rank a page's likelihood of being written by AI, like an Earwig's Copyvio Detector for LLM-written articles, and maybe integrated into the NPP tools.

2. Greater familiarity of editors with LLM-written content, both good and bad.

3. Overwhelming of the existing systems of dealing with it.

Numbers 1 and 2 need to happen before we can implement a new CSD criteria, whilst it probably won't happen until number 3. However, 3 is likely to happen either before or very soon after number 2. Mako001 (C)  (T)  🇺🇦 12:23, 12 April 2023 (UTC)

I ran across something like this recently, and I think it's worth asking. If a bot owner specifically requests a page be deleted that their bot created and there are no other edits on the page, is that page valid for WP:G7 speedy deletion? Steel1943 (talk) 07:56, 20 April 2023 (UTC)

From a BAG perspective, the bot owner is responsible for the edits made by the bot, and thus bot edits are an extension of that user. I would have no issue G7'ing a page created by a bot, assuming the request came from the operator. Primefac (talk) 08:08, 20 April 2023 (UTC)

Deletion of recently bot-created pages is fine under either G6 (usually these are results of errors) or G7 if the operator asks. Old pages may need to be retained and shouldn't just be deleted when the bot operator ragequits (but for that, it doesn't matter whether the request is made from the bot account or the operator account either). —Kusma (talk) 08:12, 20 April 2023 (UTC)

I agree with Primefac, and I think this has come up before with the same answer. I know at RfD that discussions of bot-created redirects are speedily closed as delete if the bot operator agrees and nobody else has disagreed, and G7 has been cited in at least some of those cases. Thryduulf (talk) 09:31, 20 April 2023 (UTC)

Yes. SmokeyJoe (talk) 10:10, 20 April 2023 (UTC)

I think that treating a bot as an extension of its operator is a reasonable interpretation for G7. So yes, I'd consider this a legitimate deletion. Jo-Jo Eumerus (talk) 10:20, 20 April 2023 (UTC)

An edge case that's maybe worth thinking about - what if the bot's acting on someone else's input? Who, for example, could legitimately tag this page as a G7 - Pppery, SD0001, either, neither, or only if both agree? How about archive bots? —Cryptic 11:38, 20 April 2023 (UTC)

We need keep insisting that admins have good judgement and sense.

If a bot owner started doing such things, it’s a clear flag to review their status as a bot owner. Wayward bot owners could do far more damage than a mere admin. SmokeyJoe (talk) 12:39, 20 April 2023 (UTC)

In the case of the specific page, by the letter of the criterion it cannot be correctly G7 deleted as there have been significant edits by both Ppppery and SD0001's bot. However the intent of G7 is clearly met if all authors of a page agree, so if a tag placed by the others was explicitly endorsed by the others it could be deleted, although in this case I'd rather the history be kept. Thryduulf (talk) 13:03, 20 April 2023 (UTC)

It was my understanding prior to this discussion that I could tag Wikipedia:Database reports/Possibly out-of-process deletions for G7 if I wanted to, and that SD0001 couldn't. But after thinking about it, there was no good reason for me to think that, especially since at least one person other than me has caused the bot to update that page.

I have no plans to delete it, so this is moot, and this is probably rare enough it doesn't need to be handled explicitly. * Pppery * _{it has begun...} 17:52, 20 April 2023 (UTC)

Related: should G6 be clarified that it is not to be used to delete unapproved drafts?

This is regarding Pppery's point above, observing several admins using G6 as justification to delete articles in basically the same situation as above, which is pretty clearly not uncontroversial. Is it worth clarifying somehow that this is not an approved use of G6? I don't think it's that big of a deal personally, but I'm aware of at least one high-profile Arbcom case that led to a desysop primarily on the basis of out-of-process speedy deletions. Or maybe there needs to be a blurb added to the description of the criterion reinforcing that it's not WP:IAR deletion, a page has to meet the stated conditions just like all the other criteria? I'm not proposing anything here, just interested in opinions. Ivanvector (^Talk/_Edits) 01:14, 20 April 2023 (UTC)

Until more people start getting desysopped, they're going to continue to treat it like "whatever I want to delete", no matter what this page explicitly says aren't speedies. Just look at how many U5s conveniently ignore everything after the "webhost" section title. —Cryptic 01:45, 20 April 2023 (UTC)

Yes, the amount of drafts that are deleted under WP:U5 is crazy - granted most of them are probably WP:G11. Galobtter (talk) 02:22, 20 April 2023 (UTC)

I looked at a few of the draft G6s, and in general it looked like COI cases where AfC should indeed have not been bypassed; might've been more compliant to redraftify and redirect to the existing draft but I don't necessarily oppose those specific deletions. I think part of issue is that AfC is for both non-confirmed and COI/PAID editors to submit drafts. I wouldn't necessarily mind a criterion that only applied to editors who shouldn't be bypassing AfC, but that might be harder to define as a CSD criterion. Galobtter (talk) 02:39, 20 April 2023 (UTC)

Doing something like that might also intensify the already-frequent (and frequently justified) assumptions of bad faith on the part of new-page patrollers: a new editor wrote a promotional draft, so obviously they must have an undeclared COI. —David Eppstein (talk) 05:11, 20 April 2023 (UTC)

I don't oppose this, but I don't think it will help. The things that will help are replacing G6 with a set of objective criteria (c.f #Proposal: C4 Empty maintenance categories), taking a much harder line on admins who speedy delete things out of process (up to and including desysoppings for recidivists and egregious errors), and fixing AfC so the wait time is at most a few days and only get declined when they would be speedily deletable as an article (I don't have any good ideas how to achieve fixing AfC though). Thryduulf (talk) 09:27, 20 April 2023 (UTC)

I had that thought too, that this is a symptom of the ongoing issues with G6, and that perhaps it's time to deprecate it and replace it with specific criteria for the (currently looks like four) specific situations that the policy allows. We already broke out G14, there's a proposal on this page to make "empty dated maintenance categories" a separate criterion, and "pages unambiguously created in error" easily fall under G2, G7, or R3, depending on the circumstances. That would leave G6 as a simple narrow criterion for pages blocking page moves, and I think that's a good thing. Ivanvector (^Talk/_Edits) 12:07, 20 April 2023 (UTC)

I'm not sure all "pages unambiguously created in error" easily fall under G2, G7, or R3. Back in March I G6'd approximately 22 pages (exactly half of all the G6 deletions I've ever carried out) that were supposed archives of a talk page created by Lowercase sigmabot III. The person who tagged the pages set up the archiving, and made an error in doing so, but I don't think I'd have been comfortable with using G7 on those. ~ ONUnicorn^{(Talk|Contribs)}problem solving 19:21, 20 April 2023 (UTC)

Yeah, there are unambiguous errors that are not covered under other speedy criteria. Typos not spotted for several years appear at RfD every so often for example, often the result of someone (not always the original author) moving the page very shortly after creation and leaving the redirect in place. Thryduulf (talk) 20:18, 20 April 2023 (UTC)

I agree with what I think Cryptic is saying, which is that it's the sysops, not the criteria, that are the issue here. G6 already doesn't apply to intentional things like this, so the solution to bad G6s is to complain to the admin and, if that doesn't work, go to DRV. We're playing whack-a-mole if we just rejigger the criteria: even if we can convince people that this isn't an appropriate use of G6, we'll just end up with more bad A10s or R2s or deletions that don't cite a criterion at all. It's a people problem, not a rules problem. Extraordinary Writ (talk) 20:17, 20 April 2023 (UTC)

Well, I will probably take this discussion as an inspiration to start complaining if new cases that would have been A12 come up in my database report. But I seem to be the only one willing to actually do that, despite so many people saying it should be done. * Pppery * _{it has begun...} 20:48, 20 April 2023 (UTC)

So for WP:R4 it mentions that its for redirects listed in the mainspace that redirects to select spaces like drafts, but what if its the other way around where the redirect is from a draft article that moved to the mainspace, ex. Draft:Five Nights at Freddy's (film) to Five Nights at Freddy's (film)? IAmNMFlores (talk) 21:05, 23 April 2023 (UTC)

It sounds like you're describing R2 not R4, but regardless, there's no problem with redirects from draftspace to mainspace, and they can be useful so as not to break links to the article in discussions about it prior to promotion to mainspace. The problem with main-->Draft redirects is that this would effectively circumvent the nominally unpublished status of Draft space materials (i.e. the idea is that we do not want a reader looking for an article to stumble on a draft). _signed, Rosguill ^talk 21:30, 23 April 2023 (UTC)

Redirs from draft to main usually exist because a draft passed its WP:AFC and so was moved to main. These redirects are both commonplace and harmless. --Redrose64 🌹 (talk) 22:13, 23 April 2023 (UTC)

There is consensus that redirects from draft to mainspace that result from page moves should not be deleted, see WP:RDRAFT. Other draft to mainspace redirects that get nominated at RfD are usually kept, but there are exceptions if, for example, there is no connection between the draft page's title and/or former content and the target and there is no more appropriate target. Thryduulf (talk) 08:14, 24 April 2023 (UTC)[