Commons talk:Criteria for speedy deletion
| This talk page is automatically archived by ArchiveBot. Any sections older than 60 days are automatically archived. Sections without timestamps are not archived. |
Content created by a sockpuppet
Why don't we have a Commons version of WP:G5? BodhiHarp (talk) 01:58, 5 November 2025 (UTC)
- As far as I'm aware, Commons doesn't have topic bans, general sanctions, or arbitration committees. Because Commons and Wikipedia policy don't match one another 1:1, it's not surprising that deletion policy differs as well. If there is consensus on Commons to introduce new additions to deletion policy, there's no firm barrier preventing this, but there isn't any need to specifically chase after what Wikipedia is doing either. Just because Wikipedia does something, that doesn't mean that Wiktionary, Wikisource, or Commons need to do that same thing too. --benlisquareTalk•Contribs 02:50, 5 November 2025 (UTC)
- Some admins, including myself, will use CSD G3 to wipe any files that aren't in use. However, it's worth noting that the users I do this to are usually hoaxers - we've got a couple of people that like uploading fake flags (plus one that does road signs and one that does radio stations), pretending they're real, and trying to get them into articles. The rest are serial copyvio uploaders, where we could just as easily wipe with CSD F1, or spammers where G10 applies. The Squirrel Conspiracy (talk) 04:08, 5 November 2025 (UTC)
- It's worth noting that en:Category:Wikipedia sockpuppets has about 10x as many sockpuppeteers as Category:Sockpuppeteers. It's just less of a problem here than there. Apocheir (talk) 04:27, 5 November 2025 (UTC)
- Some past discussions related to this at:
- I'm still surprised at how relaxed Commons can be about block evasion, when the original block is for copyvio uploads or hoaxes, and a confirmed sock's uploads are below the obviousness threshold for F10 or G3 (like this slow DR on whether we should delete or maybe keep the uploads from accounts that were confirmed and blocked as socks of a bad faith copyvio LTA). Belbury (talk) 11:06, 27 November 2025 (UTC)
- For hoaxers and for people blocked for willful copyvios, I agree, and maybe we should have a speedy criterion for that. But I'd hate to see us speedy-delete pictures from (say) the Colombian presidency because they were uploaded by someone who was blocked for incivility. - Jmabel ! talk 22:39, 27 November 2025 (UTC)
- Maybe we could introduce a new criterion G12 to delete content created by a sockpuppet, depending on the situation. BodhiHarp (talk) 22:52, 27 November 2025 (UTC)
- WP:G5 at Wikipedia is for content created
in violation of their ban or block
, so wouldn't affect older content from before their first block. Under an equivalent G5 we wouldn't delete content uploaded by the original uncivil Colombian presidency account, but we would delete any uploads from a later sockpuppet account that was deliberately evading that block. - I'd support a more limited Commons speedy criteria of something like
G12. Creations by banned or blocked users
Pages or files created by sockpuppets of banned or blocked users, where the user has been blocked for uploading copyrighted or hoax content.- ... Belbury (talk) 21:39, 28 January 2026 (UTC)
- Perhaps there has to be doubt on the file's copyright status or whether it is in scope for example. BodhiHarp (talk) 21:45, 28 January 2026 (UTC)
- The doubt would be in the mind of the user who decided to tag the file as G12. They wouldn't need to articulate that doubt, in the same way that en:WP:BE gives freedom for editors
to revert any edits made in violation of a ban or block, without giving any further reason
. If a copyvio sockpuppet uploaded something that was indisputably public domain and genuine, and nobody minded, it might not get tagged. Belbury (talk) 09:51, 29 January 2026 (UTC)
- The doubt would be in the mind of the user who decided to tag the file as G12. They wouldn't need to articulate that doubt, in the same way that en:WP:BE gives freedom for editors
- Not for single banned or blocked users, but for LTAs with many sockpuppets, yes. I do not delete files for a few socks, but for LTAs with a farm of socks, speedy deleting everything is the best solution. I would say OK to delete if there are more than 5 socks. Yann (talk) 22:11, 28 January 2026 (UTC)
- Two recent cases where this applies: Commons:Requests for checkuser/Case/Samathaishwarya and Commons:Requests for checkuser/Case/Universalnews23. Yann (talk) 10:42, 29 January 2026 (UTC)
- Are there many cases that go the other way, where a user blocked for copyvios or hoaxes is found to be evading that block with fewer than 5 socks, and we decide not to delete those socks' uploads? Belbury (talk) 13:29, 29 January 2026 (UTC)
- Well, I don't know, but with only a few socks, deletion should be decided on a case-by-case basis, and may need a regular DR. I can imagine a case of a user with 2 socks, and some useful uploads. Yann (talk) 14:22, 29 January 2026 (UTC)
- Speedy deletion is case-by-case, though, there's a human in the loop choosing whether or not to add the template.
- I can imagine (and have seen) cases where a user has dozens of hoax/copyvio socks but occasionally uploads something that's public domain and not faked. The question is whether it's a good use of Commons' time to carefully verify those uploads, and in subtler cases that aren't obvious enough for F1 or G3, to potentially continue hosting them for a few months while a DR is held.
- The cost of deleting a useful upload seems relatively small: if any other user notices this and cares about it, they can reupload the content or request undeletion. The cost of hosting copyvio or hoax content for a few months is potentially very large. Belbury (talk) 10:45, 12 May 2026 (UTC)
- Well, I don't know, but with only a few socks, deletion should be decided on a case-by-case basis, and may need a regular DR. I can imagine a case of a user with 2 socks, and some useful uploads. Yann (talk) 14:22, 29 January 2026 (UTC)
- Are there many cases that go the other way, where a user blocked for copyvios or hoaxes is found to be evading that block with fewer than 5 socks, and we decide not to delete those socks' uploads? Belbury (talk) 13:29, 29 January 2026 (UTC)
- Two recent cases where this applies: Commons:Requests for checkuser/Case/Samathaishwarya and Commons:Requests for checkuser/Case/Universalnews23. Yann (talk) 10:42, 29 January 2026 (UTC)
- Perhaps there has to be doubt on the file's copyright status or whether it is in scope for example. BodhiHarp (talk) 21:45, 28 January 2026 (UTC)
- For hoaxers and for people blocked for willful copyvios, I agree, and maybe we should have a speedy criterion for that. But I'd hate to see us speedy-delete pictures from (say) the Colombian presidency because they were uploaded by someone who was blocked for incivility. - Jmabel ! talk 22:39, 27 November 2025 (UTC)
Oppose - oppose automatic "damnatio memoriae" - it's a dumb/bad policy @ wp, no need to repeat it here @ wmc. it should be handled case by case/item by item, on the merits of each file &/or edit. Lx 121 (talk) 02:34, 16 July 2026 (UTC)
New Criterion: AI-generated fake
We're currently flooded with AI-generated images. All these images are generally unusable for educational purposes. I see only two uses for such files: in articles about AI and, in a very limited way, as user personal files (greetins, etc.).
Unfortunately, i very often see pseudo-historical AI-generated images in the DR nominations now, which imitate medieval pictoral arts, sculptures, documents, etc.
I propose creating a new section, F##: AI-generated fake, for quick removal such files by administrators. -- George Chernilevsky talk 10:18, 27 November 2025 (UTC)
- Agree completely. Herby talk thyme 10:52, 27 November 2025 (UTC)
Support Agreed. Yann (talk) 15:23, 27 November 2025 (UTC)- Note that Category:AI-generated images of historical figures and Category:AI-generated images of historical events both have a significant number of images in use, on Wikipedia projects that possibly don't object to AI generated content.
- What's your intended cut-off for "historical"? Would it include pictures in Category:AI-generated images of real people like a 1960s Elvis and a 2023 Yevgeny Prigozhin, where they look superficially like photographs from the period? Belbury (talk) 18:44, 27 November 2025 (UTC)
- This confirms my proposal. I see it used on the user pages and in the article en:Deepfake about AI use. Very limited use. Wikipedia projects are factual representations, not anonymous AI fantasies. I trust that administrators will use common sense when making decisions on a case-by-case basis. However, it's time for us to take action to combat this onslaught of fake images.
- For example, File:Promises of fair Compensation were often ignored or violated.jpg from Category:AI-generated images of historical events. This is a good candidate for quick deletion. The person in the foreground has shoes on one foot only. I doubt the historical resemblance of his clothing to the real thing. The two people on the left resemble cyclopes. The person on the right has three arms. Do we really need this garbage? It seems like we're moving from the world of real facts into the fake AI world of the Matrix. -- George Chernilevsky talk 20:15, 27 November 2025 (UTC)
- File:Promises of fair Compensation were often ignored or violated.jpg is an AI-generated line drawing in a generically modern style. Are you going further than
pseudo-historical AI-generated images [...] which imitate medieval pictoral arts, sculptures, documents, etc
and suggesting that any AI-generated representation of anything from history should be speedily deleted? Belbury (talk) 11:35, 26 January 2026 (UTC)- Not at all happy about the extra arm, etc., but I don't think we would be doing ourselves any favors by speedying something like this. Obviously it's an illustration, and (except for that arm) about as plausible a one as would be likely to be drawn by a human; someone could easily edit it to deal with the anomaly of the arm. - Jmabel ! talk 22:08, 26 January 2026 (UTC)
- File:Promises of fair Compensation were often ignored or violated.jpg is an AI-generated line drawing in a generically modern style. Are you going further than
Support The Squirrel Conspiracy (talk) 20:49, 27 November 2025 (UTC)
Support, but what does this accomplish that F10 doesn't already (as per Commons talk:Criteria for speedy deletion/Archive 2 § Explicitly expand F10 to include AI?)? Apocheir (talk) 01:30, 28 November 2025 (UTC)
- I believe AI should be considered as a separate criterion. It's not always obvious that it's a personal image created by AI. However, it's almost always obvious that what we're looking at is an AI-generated fake. And і don't remember AI files being DR nominated/deleted as CSD F10. F10 is usually used for selfies etc. A separate criterion will clarify and simplify the rules. -- George Chernilevsky talk 06:49, 28 November 2025 (UTC)
- @Apocheir: because these aren't ordinarily intended as "personal". - Jmabel ! talk 06:53, 28 November 2025 (UTC)
Comment while I am generally against AI-generated images being used here on Commons, I think we do have to consider that (1) several of the Wikipedias seem to consider them often acceptable, and part of our role is to support those various Wikipedias and (2) there are subjects where any visual representation will be fictional, and I'm not sure we can say "those should not be illustrated" or "any illustration of this must be created entirely be a human." For example (on point 2), if someone wanted to illustrate the fiction about George Washington chopping down a cherry tree, or the possibility that William Shakespeare had been to Italy, it is not clear to me that an AI-generated image is any less appropriate than a user-made drawing. Of course, it needs to be correctly identified for what it is. - Jmabel ! talk 06:53, 28 November 2025 (UTC)
several of the Wikipedias seem to consider them often acceptable
- out of interest, is there any interwiki project that could reach out and tell us what AI stances different projects are taking? When I see AI images listed as COM:INUSE on other projects I'm never sure if that's an endorsement or just an obscure page that hasn't been reverted yet. Belbury (talk) 09:47, 28 November 2025 (UTC)
Oppose Not needed – redundant per F10 and could cause problems of people reducing things to be a "fake" of something when that element is just intentionally an element of the image and not meant to be "faked". Prototyperspective (talk) 00:42, 2 December 2025 (UTC)
Support - was literally about to ask that question here, but add it to the already existing F10 criteria, only thing worse than AI generated image is AI Manipulated image sometimes people use to make images clearer but which intentionally makes the image less original due to random words written on the image and even making the face of the person look like someone else..that to me is a bigger problem.. Stemoc 04:05, 26 January 2026 (UTC)
Support --It's moon (talk) 08:24, 26 January 2026 (UTC)
Neutral I believe that all images created by AI are fake, and I'm not sure whether images created by humans are fake or not. But I support reconsidering the F10 criteria; has anyone nominated (SD) them yet? --Henrydat (talk) 17:54, 27 January 2026 (UTC)
Oppose This seems contrary to the purpose of speedy deletion, and likely to go wrong as a result.- Speedy deletion is not here to be 'speedy'. It is not 'urgent deletion', it is not 'important deletion' it is certainly not 'I am outraged by this' deletion. It is at best 'simple deletion', for use when the conditions are something like obvious copyright breaches or vandalism and this is so self-evident that no reasonable editor would be expected to disagree. If it's any more complicated than this, we can't use it.
- This is redundant per G1 / G3 / F10. We don't need a specific criterion to apply to a specific type of content. Criteria depend on the reason, not the medium. If it's not meeting G1 / G3, then it might still warrant deletion, but it's not straightforward enough that we can shortcircuit the discussion process and do it as a speedy.
- This is going to be used instead as a simpler way of the current crop of DRs under "Unused AI image", neither of which are valid criteria. Andy Dingley (talk) 17:55, 19 March 2026 (UTC)
Support. — 🇺🇦Jeff G. ツ please ping or talk to me🇺🇦 17:44, 19 March 2026 (UTC)
Support any way to get rid of more AI slop and faster is fine by me. Dronebogus (talk) 22:59, 30 July 2026 (UTC)
To summarize: 8 support, 3 oppose, 2 neutral.
- We are truly flooded with a ton of AI-generated garbage.
- I'm changing the rules description. I'm moving the AI rule from section F10 to a separate section, F11.
- I made more neutral description of the reason: AI-generated image without explicit educational purpose.
- Anyone can revert my changes if they believe this discussion should continue. -- George Chernilevsky talk 20:27, 14 July 2026 (UTC)
- A policy rule (especially for speedy!) that gives a vague and previously undiscussed set of reasons, then ends with "etc." as well is a recipe for chaos. This is just IDONTLIKEIT for AI. Andy Dingley (talk) 23:22, 14 July 2026 (UTC)
- The set of reasons were not written out explicitly, but they are present in the discussion at the beginning, since i was considering AI slop using real examples. "etc." removed -- George Chernilevsky talk 06:41, 15 July 2026 (UTC)
- I was apparently not counted because I failed to use the word "oppose". "I don't think we would be doing ourselves any favors by speedying something like this" was pretty clearly opposition. - Jmabel ! talk 23:44, 14 July 2026 (UTC)
- Fixed, sorry about that -- George Chernilevsky talk 06:32, 15 July 2026 (UTC)
- I'm not against this proposal, but I agree with Andy Dingley above. Speedy criteria and its description should be as clear and unambiguous as possible, but the current wording and description are just the opposite of that. Also, it seems to me the current wording "
AI-generated image without explicit educational purpose
" is way broader than what was proposed above "AI-generated fake
". Thanks. Tvpuppy (talk) 00:31, 15 July 2026 (UTC)- Changed to "AI-generated fake" -- George Chernilevsky talk 06:49, 15 July 2026 (UTC)
- A policy rule (especially for speedy!) that gives a vague and previously undiscussed set of reasons, then ends with "etc." as well is a recipe for chaos. This is just IDONTLIKEIT for AI. Andy Dingley (talk) 23:22, 14 July 2026 (UTC)
- I'll change to support for AI-generated image without
expliciteducational purpose. Don't serve an educational purpose is enough to deletion, plus AI-generated it is enough to have separate criteria. What does "fake" mean? Not-real or like-real? I believe an image is either deleted because of its content or simply because of its poor quality. If this is unclear, a discussion is necessary. --Henry (dat) 09:26, 15 July 2026 (UTC)- But this is not deletion criteria, they are speedy deletion criteria. They have to be clear and unambiguous, to the point that reasonable and experienced editors can't be expected to disagree: otherwise, if we have to start putting subjective interpretation of them into play, then they might still be good DR reasons, but they're no longer appropriate for speedy. Andy Dingley (talk) 10:14, 15 July 2026 (UTC)
- I've passed the sentence: Unless all of these are unclear. New speedy delete criteria, if it does, will allow sysops to delete files without DR nomination, or even without a speedy delete nomination, this fulfills the requirement of the nominator of this proposal. If this doesn't happen, then labeling them is necessary. --Henry (dat) 11:16, 15 July 2026 (UTC)
- I agree "AI fake" is very ambiguous. Using "without educational purpose" is definitely more specific and objective. It's moon (talk) 10:43, 15 July 2026 (UTC)
- What would be an example of a file created by generative AI that is not "fake"? - Jmabel ! talk 19:13, 15 July 2026 (UTC)
- Everyone here: Commons:Deletion requests/File:Batalla de l'Aínsa.png was insistent that it be kept. And there's nothing more real than a miracle. Andy Dingley (talk) 21:41, 15 July 2026 (UTC)
- @Andy Dingley: I can see that you are playing with irony there, but I can't tell where you actually come down on this. - Jmabel ! talk 23:20, 15 July 2026 (UTC)
- My main concern is about abuse of speedy, as is already popular without AI.
- What is a 'fake' ? The creation ab vacuo of an image of an event that is substantially held to be impossible and to have never happened would seem like one example of that. Yet when that went to DR, the !vote was strongly to keep it. This isn't about my view, it's about a collective opinion that Commons has already expressed. Andy Dingley (talk) 23:48, 15 July 2026 (UTC)
- @Andy Dingley: I can see that you are playing with irony there, but I can't tell where you actually come down on this. - Jmabel ! talk 23:20, 15 July 2026 (UTC)
- It depends on your definition of fake, as you can see from Andy Dingley's reply. I personally prefer terms like "AI fabricated" or "AI generated", which leave less room for interpretation and I'd also support over "AI fake". It's moon (talk) 22:15, 15 July 2026 (UTC)
- @It's moon: So do I understand you to be saying that this means that absolutely anything created by generative AI can be speedy-deleted? - Jmabel ! talk 23:20, 15 July 2026 (UTC)
- Sorry, I meant to support using the terms "fabricated" or "generated" over "fake", while keeping the "without educational purpose" part attached to it as @Henrydat proposed earlier. I don't support a blanket ban, but I support speedily deleting anything out of scope. It's moon (talk) 23:35, 15 July 2026 (UTC)
"absolutely anything created by generative AI can be speedy-deleted?"
- Yes, that's obviously the position that some editors here hold. This will give them the ability to do just that, without need for a DR, and without any oversight. Andy Dingley (talk) 23:44, 15 July 2026 (UTC)
- I'm sorry, I still don't get this. "Without educational purpose" is usually one of the trickier parts of a DR when that is the basis of the DR. Somehow, we seem to be saying that is magically easy when it is AI.
- I'm as opposed to AI slop as almost anyone here, short of those who think that anything touched in any manner by AI is, by definition, slop, but based, at least, on current policy (as against some harsher anti-AI policies currently under discussion) I don't see how a determination that something is out of scope because it lacks "educational purpose" is magically easier for AI-generated content than other content, so easy as to become a speedy-deletion criterion where the equivalent file, if not AI-generated, would not be eligible for speedy deletion. - Jmabel ! talk 04:02, 16 July 2026 (UTC)
- Generative AI produces a lot of low effort / low quality content en masse that can't be used for any realistic educational purpose, at a rate regular deletion requests cannot handle. When the site gets flooded at this scale a policy to speedily delete obvious out of scope files becomes necessary just to keep up.
- You are right that whether something has an educational purpose can be subjective in borderline cases, but speedy deletions should only affect obvious slop.
- Maybe we could strictly define better what an obvious case looks like, but IMO administrators can (and should) absolutely make that call. It's moon (talk) 09:08, 16 July 2026 (UTC)
- I care much less here about the issue of whether all/some AI is slop than I do about CSD being used to railroad deletions without discussion. In particular, the case you seem to be advocating here where a single admin can delete without anyone else being involved at all (there are already several admins who are not fit to be doing that).
- If we have a problem of 'too much to delete', then I still cannot see abandoning DR for CSD as being a useful fix for that. There are many much more useful things we could do first (more admins, better admins, a better workflow for DR and CfD where they don't stall as badly) rather than just giving the loudest voice and biggest advocate for deletion the ability to delete whatever they like, on sight. Andy Dingley (talk) 11:39, 16 July 2026 (UTC)
- It doesn't mean all AI-generated content has to go to delete through DR. Are there any types of files that are always deleted there? This is the subject we're talking here, or at least another topic. I know what enwiki has done with translate machine and doing with AI chatbots, unless we don't want a separate speedy delete criterion for AI-generated content. --Henry (dat) 14:32, 16 July 2026 (UTC)
- I'm looking for a typical DR example where the nominator was unaware of the AI rationale (this excludes their intention), if the result is that it was deleted by the reason "fake" --> all AI-created files must be deleted, if the result is kept because of its educational purpose --> they should never be speedy deleted. If they are deleted at CSD with the reason AI-created content or at DR with the reason fake, then our democracy will vote keep with the reason keep as an example or educational purpose. I think we're just looking for a plausible reason to delete what's going to be deleted anyway. Slop AI is easier to be delete; the fact that it's rarely used brings it closer to being out of scope. I thought I needed a support of User:Prototyperspective, they need to be put into Category:AI slop to avoid being deleted or they might even still be deleted. --Henry (dat) 19:07, 16 July 2026 (UTC)
- If I understand you correctly, Commons:Deletion requests/File:Amilamia (euskal mitologia) - Midjourney AI bertsioa.png may be an appropriate example - it's been nominated twice by different users who weren't citing any particular deletion criteria, instead saying
AI slop
(which isn't a valid reason to delete) andClear AI imagery violating Wikimedia guidelines
(where they didn't clarify which guidelines they meant). It was kept as COM:INUSE both times, and being in use is an assumed-to-be-educational angle. - There are a lot of DRs like that. Belbury (talk) 15:46, 22 July 2026 (UTC)
- If I understand you correctly, Commons:Deletion requests/File:Amilamia (euskal mitologia) - Midjourney AI bertsioa.png may be an appropriate example - it's been nominated twice by different users who weren't citing any particular deletion criteria, instead saying
- @It's moon: So do I understand you to be saying that this means that absolutely anything created by generative AI can be speedy-deleted? - Jmabel ! talk 23:20, 15 July 2026 (UTC)
- Everyone here: Commons:Deletion requests/File:Batalla de l'Aínsa.png was insistent that it be kept. And there's nothing more real than a miracle. Andy Dingley (talk) 21:41, 15 July 2026 (UTC)
- What would be an example of a file created by generative AI that is not "fake"? - Jmabel ! talk 19:13, 15 July 2026 (UTC)
- But this is not deletion criteria, they are speedy deletion criteria. They have to be clear and unambiguous, to the point that reasonable and experienced editors can't be expected to disagree: otherwise, if we have to start putting subjective interpretation of them into play, then they might still be good DR reasons, but they're no longer appropriate for speedy. Andy Dingley (talk) 10:14, 15 July 2026 (UTC)
Oppose - too broad, esp for CSD. many of those would NEED discussion to determine/decide. Lx 121 (talk) 02:31, 16 July 2026 (UTC)
Oppose the initial pseudo-historical AI-generated images
proposal, where I never got an answer on George's intended objective cut-off point for "historical". The grab bagAI-generated fake: Not realistic, fantasy, hoax, personal research, inaccurate, or pseudo-historical
that George changed it to when updating the policy a couple of days ago doesn't match the discussion that was held here: realism, fantasy, hoaxes, personal research and accuracy have not been mentioned at all. (I'd oppose that also: hoaxes are already covered by G3, personal partly by F10, everything else seems too subjective and debatable for a speedy.) --Belbury (talk) 08:22, 16 July 2026 (UTC)
Re: Criteria
'F11. AI-generated fake Not realistic, fantasy, hoax, personal research, inaccurate, or pseudo-historical.' - not ALL of those would automatically fall outside of commons' scope.
& many of the listed "reasons" are pretty vague, subjective, & would be HIGHLY debatable & require discussion, so not suitable or appropriate for CSD; could & should be discussed under PROD/deletion debates, not here.
also, the vagueness would be FAR too easy to use as deletionist excuses. Lx 121 (talk) 02:42, 16 July 2026 (UTC)
- I accept criticism, however, please offer a constructive solution. -- George Chernilevsky talk 21:06, 16 July 2026 (UTC)
Comment Perhaps I don't want to take responsibility for a new speedy criterion, but just stop for not-real AI-created content, what makes them easily recognizable. The strange thing is that the "fake" photos were kept at DR discussions, perhaps the uploaders should be held responsible for them. --Henry (dat) 01:33, 17 July 2026 (UTC)
- My constructive proposal (not a "solution", I suppose) is to revert this because there is no current consensus for something like this as a speedy deletion criterion. Most serious slop can already be speedy-deleted as G1 (nonsense). If it isn't nonsense, then I think there is no current consensus to bypass the DR process, which is the place to determine whether it is in scope as "educational" or not. I suspect that several ongoing discussions may soon come up with something stronger, but we should not jump the gun. - Jmabel ! talk 01:46, 17 July 2026 (UTC)
Wording
I have rolled back the new criterion, as there is no agreed-on wording, and it is not clear that everyone is !voting on the same thing. It looks like we have rough consensus to deal with AI slop in some way, but we need to be much more rigorous with speedy criteria so that it can be as unambiguous as possible whether files qualify or not. I think we should first refine the wording, and then have a final !vote to adopt it. -- King of ♥ ♦ ♣ ♠ 00:20, 31 July 2026 (UTC)
- Here's my attempt at making the wording more concise:
- Inaccurate AI-generated images
- To be hosted on Commons, AI-generated files must be realistically useful for an educational purpose. This criterion applies to obvious cases of low-quality or inaccurate content created with Generative AI, i.e., images featuring significant errors or fabrications such as misgenerated body parts, objects or words, as well as pseudo-historical imagery illustrating historical topics or resembling actual historical photographs, sculptures or documents.
- I was also considering adding historical subjects with obvious inaccuracies or bias regarding race or ethnicity as well as pseudo-medical imagery, e.g., images imitating medical or scientific documents, diagrams or medical imaging such as X-rays, CT scans, and MRI, but I ultimately believe these are more complex cases that should probably be left to regular deletion requests. It's moon (talk) 08:18, 31 July 2026 (UTC)
- Can you be more specific about what you mean by "pseudo-historical imagery, e.g., images imitating historical art, sculptures or documents"? Are you saying that all images of the form "X in the style of a Renaissance painting" should be eligible for speedy deletion? If so, then why does that deserve to be deleted more than "X in the style of (modern artist)?" If not, then your criterion should be reworded to remove that implication. King of ♥ ♦ ♣ ♠ 20:44, 31 July 2026 (UTC)
- The wording should mention scope, specifically that files "must be realistically useful for an educational purpose." These files are erroneous/misleading/confusing/deceptive in a way that contradicts the purpose of Commons. Apocheir (talk) 22:07, 31 July 2026 (UTC)
Done It's moon (talk) 15:26, 1 August 2026 (UTC)
- What's this saying - that an inaccurate AI image is not considered to be educationally useful, or that (like the clause in COM:G10 for adverts) such an image should only be speedied if it does not otherwise seem educationally useful? Belbury (talk) 15:46, 1 August 2026 (UTC)
- Both? We already have an exception for AI images used to educate about AI. For example, File:AI generated figure published in a Frontiers journal.png is in scope as an infamous example of an inaccurate AI illustration from medical research. If someone uploads another AI medical illustration with errors, and there's no other reason to keep it, that file would be eligible for speedy deletion.
- Keep in mind that admins have discretion with speedy deletion, and any unclear cases can go through the regular deletion process. Apocheir (talk) 18:46, 1 August 2026 (UTC)
- What's this saying - that an inaccurate AI image is not considered to be educationally useful, or that (like the clause in COM:G10 for adverts) such an image should only be speedied if it does not otherwise seem educationally useful? Belbury (talk) 15:46, 1 August 2026 (UTC)
Thank you, King of Hearts, for your participation in this topic.
I am not a native English speaker, so I may miss some subtleties of the wording.
I'm not opposed to AI; on the contrary, I use AI for software development. It's like having a not-so-precise but fast partner who suggests solutions and also handles many routine tasks. However, the corrections of mistakes and use of this code are my responsibility.
To understand the limits of reasonable AI applications, I propose creating two lists here: acceptable and unacceptable areas of AI application. I look forward to additional contributions and discussion of the elements of this list here. -- George Chernilevsky talk 23:25, 31 July 2026 (UTC)
Acceptabe
- To illustrate articles about AI
- To create advertising posters and banners for Wiki events, such as those dedicated to Wikimania, if regular photographs are not sufficient
- To illustrate articles about scientific humor like Russell's teapot.
- To illustrate articles about fairy tales, if human-created illustrations are not available.
- For seasonal greetings in user spaces.
-- George Chernilevsky talk 23:25, 31 July 2026 (UTC)
Unacceptabe
- All biology, including anatomy, microbiology, botany, paleontology, cryptozoology, and infographics on biology. Any inaccuracy here is unscientific.
- Food and drink, household items, clothing, and footwear. We have a plenty of real photographs of these items.
- All real industrial objects, including all types of transport, engines, electonic, machine tools, and complex mechanisms.
- Existing or previously existing architecture. Real drawings, paintings, or photographs should be used, rather than straying from reality.
- Religion. Images of deities, demigods, heroes, and demons are canonized in most cases. I would also add here the heroic epics of various peoples as religion for the same reason. Although religion is not a pure science, inaccurate AI can offend the sensibilities of believers. And I don't think we need an "icon" created by AI. Only photographs of actual paintings or sculptures should be used.
- Pseudo-historical reconstructions. Depictions of historical events should be created by people with in-depth knowledge of the subject. AI cannot create a reconstruction that is consistent with existing knowledge.I've looked at thousands of AI images of historical events, and they all contain a lot of errors. I am not a historian, but the errors are obvious.
- Imitation of an artistic style. AI images that resemble a painting and the artist's style, or an old document or engraving on yellowed paper, or imitations of an old photograph. These are obvious fakes, dangerous because they can be perceived by the viewer as real works of art.
-- George Chernilevsky talk 23:25, 31 July 2026 (UTC)
Oppose - your revised blanket ban is even worse! in every single one of the categories you listed, there are clear & obvious uses for ai created/edited material WITHIN commons' scope.(your rationales also open the "back" door to banning non-ai instances of same, which is deeply problematic in its own ways...) & several of your rationales clearly fail commons' policy (not offending the sensibilities of believers, "we have plenty of", "we don't need" ) & pretty much EVERY circumstance you described would be a matter for rfd discussion, NOT csd. Lx 121 (talk) 18:34, 7 August 2026 (UTC)
Resuming discussion
This is getting very far afield of a discussion about speedy deletion criteria. We have had (and are having) lengthy discussions elsewhere about what are and are not acceptable uses of AI. Please let's not have yet another general discussion in parallel with those. - Jmabel ! talk 00:21, 1 August 2026 (UTC)
Agree - & csd is STILL not the place to settle AI-content policy. if anything, policy changes HERE, should happen after we reach a community consensus on the general issues. Lx 121 (talk) 18:25, 7 August 2026 (UTC)- @Jmabel and @Lx 121: Could you clarify if your feedback is meant to address my proposal, the latest one by George Chernilevsky, or both? Thanks! It's moon (talk) 20:20, 7 August 2026 (UTC)
- @It's moon: Here, I was mostly replying to George Chernilevsky's lengthy list of criteria. More generally, though, I think it is premature to add something to the list of speedy deletion criteria while we are currently very much in the midst of deciding what generative-AI content is acceptable for Commons in the first place, let alone whether it can be speedied vs. needing a full DR. - Jmabel ! talk 22:40, 7 August 2026 (UTC)
Criteria for templates
Does the criteria for templates also apply to modules? BodhiHarp (talk) 05:20, 28 January 2026 (UTC)
- @BodhiHarp: I'd hesitate to speedy-delete a module except on the usual very general criteria like it being vandalism/nonsense, or if it were unused and the request came from the creator. The two speedy criteria specifically for templates are "unused" and "duplicate." If a module were unused, I'd probably still like to understand why before deleting it; if it were a duplicate, I'm not even sure how we handle that; I don't think I've seen a redirected module, not sure that can even be done. Certainly more of a process than I'd like to see handled as a "speedy".
- Do you have a specific case (past or present) in mind, or was this an abstract question? I'm not sure I've ever seen a module deleted. - Jmabel ! talk 19:27, 28 January 2026 (UTC)
- Module:IPA symbol/data is unused and is a mirror of a module on Wikipedia. BodhiHarp (talk) 19:30, 28 January 2026 (UTC)
Proposal to change two criteria
I propose this:
- To change G2 so the broken redirect part is a part of G8, and then rename it to R1
- To merge G5 into G6
BodhiHarp (talk) 03:49, 16 February 2026 (UTC)
- @BodhiHarp Would you please include proposed changes in language? — 🇺🇦Jeff G. ツ please ping or talk to me🇺🇦 08:26, 16 February 2026 (UTC)
- Changing G2
- I am proposing to change G2. Unused and implausible, or broken redirect so the broken redirect part is a part of G8. Page dependent on deleted or non-existent content, and then rename it to R1. Unused and implausible redirect, which means we should move G2. Unused and implausible, or broken redirect into a new #Obsolete section with the same old wording and rule:
- G2. Unused and implausible, or broken redirect
- Page is an unused and implausible redirect, or a redirect that is dependent on deleted or non-existent content. Unused talk page redirects created as a result of a page move and cross-namespace redirects may also be deleted under this criterion.
- change G8. Page dependent on deleted or non-existent content so the broken redirect part is a part of it:
- G8. Page dependent on deleted or non-existent content
- The page or file depends on content that was deleted or no longer existing, such as orphaned talk pages without useful content, subpages without parent page, redirects that are dependent on deleted or non-existent content, and so on. The criterion only applies to content within Wikimedia, and does not apply to external content (i.e., deleted source).
- and then finally renaming G2 to R1. Unused and implausible redirect:
- R1. Unused and implausible redirect
- Page is an unused and implausible redirect. Unused talk page redirects created as a result of a page move and cross-namespace redirects may also be deleted under this criterion.
- Merging G5
- I am proposing to simply merge G5. Temporary deletion for history cleaning or revision suppression into G6. Uncontroversial maintenance:
- G6. Uncontroversial maintenance
- Content temporarily deleted to make way for a page move, content temporarily deleted to remove specific revisions containing vandalism, threats/attacks, or personal information, content temporarily deleted to perform a history merge or split, or other uncontroversial maintenance tasks that require temporary or permanent deletion.
- and then move G5 to the same new obsolete section with the old wording and rule:
- G5. Temporary deletion for history cleaning or revision suppression
- Content temporarily deleted to remove specific revisions containing vandalism, threats/attacks, or personal information. Content may also be temporarily deleted to perform a history merge or split.
- BodhiHarp (talk) 03:51, 17 February 2026 (UTC)
G7 2026 discussion
The original discussion that decided 7 days for G7 is Commons_talk:Criteria_for_speedy_deletion/Archive_1#c-NickK-2011-04-07T19:52:00.000Z-Author_or_uploader_request_deletion_of_recently-created_unused_page_or_file.
Commons:Village pump/Archive/2026/02#Courtesy deletion requests by uploaders, aka CSD G7 this discussion showed that com:courtesy deletion requests (aka CSD G7) would be declined if the files are com:inuse, and that the period of grace would remain at 7 days, but some users were open to extending it to 2 weeks. RoyZuo (talk) 13:53, 5 March 2026 (UTC)
Move
I suggest moving this page to Commons:Speedy deletion for the same reason as w:Wikipedia talk:Speedy deletion/Archive 90#Requested move 19 January 2025, I will also notify the village pump of this. Crouch, Swale (talk) 10:25, 18 April 2026 (UTC)
- Why does it matter which direction the redirect goes? - Jmabel ! talk 18:31, 18 April 2026 (UTC)
- I don't think it matters. — 🇺🇦Jeff G. ツ please ping or talk to me🇺🇦 18:46, 18 April 2026 (UTC)
- I came here to propose it as well. Can someone kindly do it? All the reasons outlined in enwiki's proposal apply to this page as well. FaviFake (talk) 17:47, 12 May 2026 (UTC)
- Again, @FaviFake why does it matter which direction the redirect goes? - Jmabel ! talk 19:56, 12 May 2026 (UTC)
- What does the direction of the redirect have to do with this? I'm proposing the page be moved to a new title per the reasons listed in the discussion in the opening comment:
[...] I am proposing a move from Wikipedia:Criteria for speedy deletion to Wikipedia:Speedy deletion for the following reasons:
FaviFake (talk) 20:11, 12 May 2026 (UTC)- To increase emphasis on the process itself, rather than the criteria. It is evident at this point that this page isn't only about a set of criteria but also an established process to delete pages based on the criteria.
- To enable titling consistent with the other deletion process pages (Wikipedia:Articles for deletion, Wikipedia:Proposed deletion). [Here con commons, that's be Deletion requests, not Criteria for deletion —FaviFake (talk)]
- To allow smoother referencing and better syntax, e.g. "tag it for WP:Speedy deletion" vs. "tag it for WP:Criteria for speedy deletion".
- The page currently contains topics other than just the criteria, such as the step-by-step instructions, the procedure, and information about the process. A rename of this page could make room for expanding/altering the scope if needed in the future.
- @FaviFake: (1) if you use "WP" as a prefix on Commons, it doesn't work. (2) COM:Speedy deletion already does exactly what you want it to do; it just happens to do it via a redirect. - Jmabel ! talk 20:55, 12 May 2026 (UTC)
- @Jmabel I was quoting Frost's proposal on WP, so obviously the links don't work here. And I know the redirect points there, I don't want to change the various redirects. I'm merely proposing the title of this page be changed to COM:Speedy deletion. FaviFake (talk) 14:35, 13 May 2026 (UTC)
- What does the direction of the redirect have to do with this? I'm proposing the page be moved to a new title per the reasons listed in the discussion in the opening comment:
- Again, @FaviFake why does it matter which direction the redirect goes? - Jmabel ! talk 19:56, 12 May 2026 (UTC)
Proposal to add clarification about template case-sensitivity
This should be fairly uncontroversial, but I wouldn't quite describe it as a minor edit, so here goes. My proposal is simply to add the following to the end of the last paragraph of the opening section (To tag a page for speedy deletion, you may use ...
), to hopefully prevent others from making the same mistake I almost did:
(Please note that SD is case-sensitive; Sd or sd will use the unrelated template {{Sd}}.)
Wikitext, since I had this all ready to submit before I thought better of it since it does add some content to the page:
(Please note that <kbd>SD</kbd> is case-sensitive; <kbd>Sd</kbd> or <kbd>sd</kbd> will use the unrelated template {{tl|sd}}.)
Thoughts? Kinsio (talk) 19:14, 13 July 2026 (UTC)
- Only that I'm not sure why the <tvar>s are there. The contents of those are literals, and won't change with language. - Jmabel ! talk 23:14, 13 July 2026 (UTC)
- @Jmabel: If I'm understanding the documentation correctly, that's precisely the point. The terminology is a bit counter-intuitive, but I'm pretty sure it's just a variable in the broader programming sense (a value with a name associated with it). The
<tvar>s I used should make the text appear to a translator as (Please note that $speedy5 is case-sensitive; $sd1 or $sd2 will use the unrelated template $sd3.), and putting those variable names in the translated text will replace them with their corresponding values in the rendered translation. Kinsio (talk) 01:42, 14 July 2026 (UTC)
- @Jmabel: If I'm understanding the documentation correctly, that's precisely the point. The terminology is a bit counter-intuitive, but I'm pretty sure it's just a variable in the broader programming sense (a value with a name associated with it). The
Can these be removed
Unicode 1DD00-1DDFF Dingbats Supplement and Unicode 1DE00 Combining Diacritical Marks Additional? Rugio Valerio (talk) 15:29, 7 August 2026 (UTC)
- @Qawerasersa: this is a page for discussion the criteria for speedy deletion. I don't actually understand your remark (in terms of what you want done) but I suspect you made it in the wrong place, and that you were relying on context that is not here. - Jmabel ! talk 17:53, 7 August 2026 (UTC)
- also, why exactly do you want those apparent(?) unicode files removed? o__0 Lx 121 (talk) 18:14, 7 August 2026 (UTC)
- Only because those are fake unicode. Rugio Valerio (talk) 19:36, 7 August 2026 (UTC)