Wikipedia talk:Criteria for speedy deletion

Read this before proposing new or expanded criteria Shortcut WP:NEWCSD Contributors frequently propose new (or expansions of existing) criteria for speedy deletion. Please bear in mind that CSD criteria require careful wording, and in particular, need to be Objective: Most reasonable people should be able to agree whether a page meets the criterion. Often this requires making the criterion very specific. Uncontestable: It must be the case that almost all pages that could be deleted using the rule, should be deleted, according to consensus. CSD criteria should cover only situations where there is a strong precedent for deletion. Remember that a rule may be used in a way you don't expect, unless you word it carefully. Frequent: Speedy deletion is intended primarily as a means of reducing load on other deletion methods such as Wikipedia:Articles for deletion and Wikipedia:Proposed deletion. These processes are more discriminating because they treat articles case-by-case, and involve many points of view; CSD sacrifices these advantages in favor of speed and efficiency. If a situation arises only rarely, it's probably easier, simpler, and fairer to delete it with one of the other methods instead. This also keeps CSD as simple and easy to remember as possible, and avoids instruction creep. Nonredundant: If the deletion can be accomplished using a reasonable interpretation of an existing rule, just use that. If this application of that rule is contested, consider discussing and/or clarifying it. New rules should be proposed only to cover situations that cannot be speedily deleted otherwise. If you do have a proposal that you believe passes these guidelines, please feel free to propose it on this discussion page. Be prepared to offer evidence of these points and to refine your criterion if necessary. Consider explaining how it meets these criteria when you propose it. Do not, on the other hand, add it unilaterally to the CSD page. this header: view • edit

This is the talk page for discussing improvements to the Criteria for speedy deletion page.

Put new text under old text. Click here to start a new topic. New to Wikipedia? Welcome! Learn to edit; get help. Assume good faith Be polite and avoid personal attacks Be welcoming to newcomers Seek dispute resolution if needed Shortcut WT:CSD

Archives: Index, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23, 24, 25, 26, 27, 28, 29, 30, 31, 32, 33, 34, 35, 36, 37, 38, 39, 40, 41, 42, 43, 44, 45, 46, 47, 48, 49, 50, 51, 52, 53, 54, 55, 56, 57, 58, 59, 60, 61, 62, 63, 64, 65, 66, 67, 68, 69, 70, 71, 72, 73, 74, 75, 76, 77, 78, 79, 80, 81, 82, 83, 84, 85, 86, 87, 88, 89Auto-archiving period: 30 days

Text and/or other creative content from this version of Wikipedia:Categories for discussion/Speedy/Criteria was copied or moved into Wikipedia:Criteria for speedy deletion with this edit on 20:38, 4 December 2013. The former page's history now serves to provide attribution for that content in the latter page, and it must not be deleted as long as the latter page exists.

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)[reply]

There have been a number of times where I have observed a page nominated for G10 which meets every criteria for being a G10 (attack page) except that it is sourced. Sometimes it's improperly sourced (e.g. through an external link rather than reference) and I don't think I've seen one with a high quality reliable source but a source exists. These pages normally end up deleted - and good riddance I say to them even though I'm reluctant to press the button - despite not technically qualifying under the criteria. I'm wondering if we want to soften the wording of that requirement given this practice. Best, Barkeep49 (talk) 17:16, 11 January 2023 (UTC)[reply]

I've always assumed that the unsourced bit only applies to the "biographical material about a living person that is entirely negative in tone" part, so it's fine to delete something under G10 if it's sourced if it meets any other part of G10 (e.g. "material intended purely to harass or intimidate a person"). It is entirely possible to have a legitimate encyclopedia article about a living person which is entirely negative in tone, e.g. most articles about criminals. Removing the sourcing requirement entirely would allow them to be deleted under G10. Hut 8.5 17:52, 11 January 2023 (UTC)[reply]

I agree with @Hut 8.5. Look at the grammar and punctuation of the sentence.

Examples of "attack pages" may include libel, legal threats, material intended purely to harass or intimidate a person or biographical material about a living person that is entirely negative in tone and unsourced. (emphasis added)

I think this could be clarified thusly:

Examples of "attack pages" may include: libel, legal threats, material intended purely to harass or intimidate a person, or biographical material about a living person that is entirely negative in tone and unsourced. (addition of a colon and an oxford comma.)

Either way though, I think Hut is right that the unsourced part only applies to the last item in the list. ~ ONUnicorn^{(Talk|Contribs)}problem solving 18:12, 11 January 2023 (UTC)[reply]

I agree with Hut8.5 and ONUnicorn and would support adding that punctuation for clarity. Thryduulf (talk) 19:07, 11 January 2023 (UTC)[reply]

This matches my understanding and I support the proposed punctuation change. Jclemens (talk) 19:41, 11 January 2023 (UTC)[reply]

If a new article reads "John Snotblot is a fucking asshole", and this sentence ends in 16 footnotes leading to online sources that state unanimously that he's a fucking asshole, I'm going to submit the article for G10 deletion, even though it's clear that he's widely noted for being a fucking asshole, meeting WP:N without question.

However, if it reads "John Snotblot, the president of Marvinia since January 2, 2023, is a fucking asshole", then I'm going to change it to read "John Snotblot has been the president of Marvinia since January 2, 2023", possibly keeping one of the existing sources that corrobate his presidency and the date his term began even if it does also call him a fucking asshole, or else replacing all the original sources with one neutral source. Then I'm going to leave an "only" warning on the creator's talk page and call it a day, leaving it to others to produce a respectable "Criticism" section. Largoplazo (talk) 23:07, 11 January 2023 (UTC)[reply]

Even if your example was not libellous and not intended to purely harass or intimidate the subject there is no need for G10 as it meets criteria A1 and A7. Thryduulf (talk) 23:36, 11 January 2023 (UTC)[reply]

Well, the president of Marvinia bit obviously disqualifies it from both. I think Largoplazo was referring to how they read G10, and I think their assessment is purely right. Nobody is suggesting we scrap G10. casualdejekyll 02:55, 12 January 2023 (UTC)[reply]

My reading is in line with that of ONUnicorn et al: the 'unsourced' only applies to the 'biographical material that is entirely negative in tone' bit. I'd add that this part of the text is listing examples that may qualify - it's not an exhaustive list, and it doesn't define the criterion. Any page, regardless of sourcing, that disparages, threatens, intimidates, or harasses their subject or some other entity, and serves no other purpose, can be deleted under G10. Having said all that, I would have no problem with a tweak in the wording to avoid any possible confusion on this point. Girth Summit _(blether) 18:57, 12 January 2023 (UTC)[reply]

Hut8.5 is correct. If I remember correctly, the "negative unsourced BLP" part of G10 was introduced after Wikipedia:Requests for arbitration/Badlydrawnjeff, principle 4 ("summary deletion of BLPs"). – filelakeshoe (t / c) 🐱 19:57, 13 January 2023 (UTC)[reply]

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)[reply]

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)[reply]

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)[reply]

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)[reply]

"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)[reply]

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)[reply]

• 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)[reply]

  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)[reply]

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)[reply]

​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)[reply]

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)[reply]

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)[reply]

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)[reply]

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)[reply]

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)[reply]

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)[reply]

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

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)[reply]

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)[reply]

A recent AfD, Wikipedia:Articles for deletion/Gregory Marchand, appears to raise the following issue: the same article was previously deleted and salted under different variations of the subject's name, but may not be close-enough to the deleted version for G4 to apply. G5 could potentially apply, as the previous versions also had sockpuppet issues, but the 2020 creator of the new version is as of now unblocked. Should there be a clause of G4 that applies to creation of the same topic as a deleted and delete-protected article under a different title, even when the content may differ? Or is that too much of a stretch for CSD? (Marchand appears to be headed for sure deletion regardless, but it would be helpful to be able to shortcut such discussions without having to use excuses like WP:SNOW.) —David Eppstein (talk) 08:30, 16 January 2023 (UTC)[reply]

I have observed, and even noted it once at DRV, that G4 is in practice given wider and stronger applicability when the topic’s earlier title is WP:SALTED. That is, if someone sidesteps the salting by recreating under a variant title, admins do G4 and WP:SALT the variant title, and do not get criticised for it at DRV. I think this is good practice, even if undocumented.

SALTing should be read as a prohibition on the creation of the same topic under any title. Challenges to the SALTing should be made at WP:RfPP or WP:AfC. WP:THREE is very good advice on how to make a good case. Challenging a SALTing at DRV should be reserved for complaints of denial of unsalting at RfPP. SmokeyJoe (talk) 08:47, 16 January 2023 (UTC)[reply]

G4 is only for substantially similar pages. If there is a case where content may differ then it is not eligible for G4, regardless of the location of the page, because in theory there could be new information or writing that overcame the original deletion rationale (e.g. GNG is now met). Primefac (talk) 09:41, 16 January 2023 (UTC)[reply]

To nitpick, the GNG can become met without any changes to the page, because the GNG refers to sources that exist, not sources used on the page.

For this sort of reason, I think G4s should readily slip into draftifications. If the page looks unimproved, put it in draft for the new source to be added. SmokeyJoe (talk) 03:00, 27 January 2023 (UTC)[reply]

G4 already applies to the page "having any title", so people circumventing SALT by using different titles for the same content is already covered. But, as Primefac notes, like all speedy criteria, there needs to be an objective standard. If the content is no longer fundamentally the same, it shouldn't be a single admin's decision whether the original reasons for deletion still apply (especially when the page contains information that was not available in the first XFD). Regards SoWhy 10:32, 16 January 2023 (UTC)[reply]

I support more liberal use of G4 in mainspace in the six months following an AfD consensus to delete. This would include a re-creation using different but similarly failing sources to those that didn’t satisfy the participants in the AfD. New sources, new since the AfD, not merely different sources, would amount to a sufficient difference to make G4 proper. If a proponent for re-creation of a recently deleted article disagrees, the answer is AfC, not a title variation, even if it includes a different bad source.

How long should a consensus at AfD to delete be respected, prohibiting bold re-recreation without overcoming the reasons for deletion? I think the answer is six months.

Some admins err in SALTing for too long. I think a SALTing over six months should be dependent on an AfD consensus to SALT. SmokeyJoe (talk) 11:09, 16 January 2023 (UTC)[reply]

I understand where you are coming from and I don't think G4 is misapplied if the changes were only made to try and game the system, e.g. by replacing one bad source with another bad source (call it IAR or applying the spirit of the policy). But that requires a blatant attempt to circumvent G4. If the reviewing admin has to start judging the merits of the new article, the applicability for G4 ends. And I would also oppose any kind of recreation ban without going through AFC, considering that AFC is not a requirement for new articles and AFC is completely overwhelmed already anyway. However, that is not a question for CSD but instead would require a change to the deletion policy (after a widely advertised RFC). Regards SoWhy 11:50, 16 January 2023 (UTC)[reply]

I'll take a more conservative tack than SoWhy: Substantially altering the article in any way, such as the addition of a reference that had been mentioned in the AfD discussion but not put into the article, makes G4 inapplicable. There is no time limit on G4, so theoretically if such a change were made 15 minutes after the discussion were closed, G4 would remain inapplicable. That's what G4 says.

Now, would the act of doing so be disruptive editing? Possibly, but that's a judgment call requiring a determination of user conduct. Since it's not pure vandalism, one admin acting unilaterally would be inappropriate. In such cases, the user conduct becomes the discussion: the article content, having been sufficiently changed to avoid G4, might well be deleted again later. Of course, if the user can demonstrate that the addition of content solved the problem that led to the deletion, then it's obviously no longer a conduct issue: we don't punish editors who improve the encyclopedia simply for editing with timing that looks bad. Jclemens (talk) 08:39, 17 January 2023 (UTC)[reply]

To expound a bit more on why SmokeyJoe's take on G4 is wrong, let us look no farther than the text itself: It excludes pages that are not substantially identical to the deleted version, and pages to which the reason for the deletion no longer applies. Since the sentence contains two clauses it is clear that there can by copies that are substantially identical to the deleted version but to which the reason for the deletion clearly still applies, as well as pages to which the reason for the deletion no longer applies. that are substantially identical to the deleted version. Now, I think the latter may be a null set (except, perhaps, if policy changed to make previously unacceptable content acceptable), but CSD policy is written to ensure G4'able content must be both substantially identical and the reason for the deletion still applies. Jclemens (talk) 08:47, 17 January 2023 (UTC)[reply]

I agree completely with Jclemens - both parts of G4 must be met for it to be applicable. I can also think of another example of where the page is substantially identical but the reason for deletion no longer applies - where the deletion reason was based on a premise which has since been shown to be incorrect. In every such case I can think of, and indeed every case I can think of where Jclemens' scenario would apply, going via the deleting admin and/or DRV would be better than just unilaterally recreating the article, but I'm not going to say doing so would never be correct. Thryduulf (talk) 15:56, 17 January 2023 (UTC)[reply]

I believe there may be space between policy as written and policy in practice. G4s happen where the admin thinks it would be a SNOW deletion per the recent AfD. The criteria for SALTing is lower than for G4. Also, in NPReview, a re-creation after a recent AfD is likely to be WP:DRAFTIFY-ed (fitting criterion 2a-ii). SmokeyJoe (talk) 00:17, 18 January 2023 (UTC)[reply]

I believe there may be space between policy as written and policy in practice. When it comes to CSD there isn't - the consensus going back years that

• The criteria for speedy deletion (CSD) specify the only cases in which administrators have broad consensus to bypass deletion discussion, at their discretion, and immediately delete Wikipedia pages or media. and
• Administrators should take care not to speedily delete pages or media except in the most obvious cases.

combined with the nature of speedy deletion and the deletion policy mean that pages must meet the letter and spirit of the criterion to be eligible. Thryduulf (talk) 01:37, 18 January 2023 (UTC)[reply]

Ideally, yes. Sometimes, stuff happens without “broad consensus”, despite the care that should be taken.

My reading of DRV and WT:CSD over the years is that liberal G4 deletions are negligible compared to liberal G6 deletions. SmokeyJoe (talk) 02:59, 18 January 2023 (UTC)[reply]

G6 is indeed by far the most misused speedy deletion criterion (which is why I've been proposing reform of it for years), but misuse is still misuse and should never be endorsed however frequently it does or does not happen. Basically if a page does not meet both the letter and spirit of a CSD criterion then it cannot be correctly speedily deleted. If there are pages you think should be deleted but cannot be then propose a new or modified criterion and get that consensus - c.f. WP:IARUNCOMMON. Thryduulf (talk) 10:17, 18 January 2023 (UTC)[reply]

• In order for G4 deletions to meets its letter, not just it’s spirit, I propose that it’s wording be modified. There should be mention of thresholds. The soft words are “sufficiently” and “substantially”. These should be connected to sources. Things that affect the threshold of applicability of sufficiently/substantially include: the age of the AfD (six months default threshold); and whether the title has been SALTed. At the risk of bloat, consider advice to use REFUND to draftspace and improve the draft to make the case that the reasons for deletion have been overcome. —SmokeyJoe (talk) 00:54, 27 January 2023 (UTC)[reply]

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)[reply]

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)[reply]

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)[reply]

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)[reply]

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)[reply]

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)[reply]

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)[reply]

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)[reply]

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)[reply]

@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)[reply]

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)[reply]

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)[reply]

• 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)[reply]

  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)[reply]

  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)[reply]

  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)[reply]

  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)[reply]

  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)[reply]

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

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

  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)[reply]

  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)[reply]

  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)[reply]

  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)[reply]

  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)[reply]

  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)[reply]

  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)[reply]

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)[reply]

• 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)[reply]

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)[reply]

• 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)[reply]

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)[reply]

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)[reply]

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)[reply]

@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)[reply]

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)[reply]

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)[reply]

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)[reply]

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)[reply]

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)[reply]

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)[reply]

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)[reply]

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)[reply]

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)[reply]

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)[reply]

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)[reply]

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)[reply]

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)[reply]

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)[reply]

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

• 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)[reply]
• 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)[reply]

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)[reply]

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

(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)[reply]

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)[reply]

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)[reply]

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)[reply]

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)[reply]

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)[reply]

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)[reply]

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)[reply]

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)[reply]

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)[reply]

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)[reply]

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)[reply]

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)[reply]

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

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)[reply]

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

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)[reply]

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)[reply]

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)[reply]

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)[reply]

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)[reply]

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)[reply]

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)[reply]

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)[reply]

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)[reply]

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)[reply]

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)[reply]

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)[reply]

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)[reply]

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)[reply]

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)[reply]

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)[reply]

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)[reply]

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

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)[reply]

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)[reply]

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)[reply]

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)[reply]

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)[reply]

• 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)[reply]
  • 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)[reply]
    • 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)[reply]

      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)[reply]

• 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)[reply]

  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)[reply]

  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)[reply]

  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)[reply]

  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)[reply]

  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)[reply]

• 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)[reply]

  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)[reply]

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)[reply]

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)[reply]

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)[reply]

• 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)[reply]
• 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)[reply]

• 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)[reply]

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)[reply]

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)[reply]

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)[reply]

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)[reply]

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)[reply]

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)[reply]

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

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)[reply]

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

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)[reply]

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)[reply]

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)[reply]

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)[reply]

• 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)[reply]
• 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)[reply]
• 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)[reply]

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)[reply]

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

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

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)[reply]

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)[reply]

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)[reply]

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)[reply]

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)[reply]

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)[reply]