AI on FMC - Voting {{s}} or {{o}}

(voting ends on January 11, 2026)
Voting period is over. Please don't add any new votes.

•  Allow I say it's a case-by-case situation. There are some specific AI-generated works which are notable, such as Théâtre D'opéra Spatial, but as a rule of thumb I don't expect most AI media to pass. JayCubby (talk) 16:58, 18 December 2025 (UTC)

•  Support Per JayCubby, and it depends on how the stance and progress of AI evolves in the near future --PantheraLeo1359531 😺 (talk) 17:21, 18 December 2025 (UTC)

•  Allow If some contributor uses AI tools to generate a good illustration (educationally useful, without copyright concerns, freely licensed, with correct content), then I don't see why it should be forbidden. There won't be a technical difference between using pens on paper and a scanner, a drawing tablet and some Adobe software or AI software. AI software would only reduce the need to have actual drawing abilities (and demand wording abilities). Grand-Duc (talk) 17:23, 18 December 2025 (UTC)

•  Neutral It totally depends on the complexity of creation of such media, if it can be made by anyone using a simple command and using simple tools, then anyone can create their own required media and won't come seeking them in commons I believe but, if it is a task of expertise skills then definitely it should be allowed. As I have zero experience in creation of such content, I am in no position to vote for or against it, Thank you -- Paramanu Sarkar (talk) 17:24, 18 December 2025 (UTC)

 Allow Sure, as long as that specific AI media is well known. PublicDomainFan08 (talk) 17:25, 18 December 2025 (UTC)

•  Neutral I have no experience in AI generated media. -- Tisha Mukherjee (talk) 17:36, 18 December 2025 (UTC)
•  Allow If the illustration is well done, not misleading and useful for the purpose of presenting the topic, it should be allowed. -Theklan (talk) 17:42, 18 December 2025 (UTC)
• I agree that it depends. For me, it's OK. As long as the AI image meets the FP rules. Such as resolution, size, etc. But I think we should think carefully. Because the work that a photographer does, no AI will ever replace... heylenny (^talk/_edits) 17:46, 18 December 2025 (UTC)
•  Allow per JayCubby, I also don't think most AI media would pass FM, but in my opinion a blanket ban isn't needed. Thanks. Tvpuppy (talk) 18:16, 18 December 2025 (UTC)

•  Allow No need or sufficient reason to prohibit any production-method or -tool. There may be better quality content in the future and one can also use an AI-generated image, and after changing it via further prompts, also modify it further in an image editor like GIMP so the line can be quite blurred. --Prototyperspective (talk) 18:24, 18 December 2025 (UTC)

•  Info For everyone here talking about images and illustrations: This voting is only for media, FMC (video, sound, gifs, etc.). There may be grounds to have a similar discussion and vote at FPC for images (jpg, tif, etc.). --Cart (talk) 18:39, 18 December 2025 (UTC)

  Thanks for clarifying. One can use such images for videos and gifs and the same can also be done for other types of media and I should have used the term "media" instead of images (e.g. editing videos via prompts and a video editor). Prototyperspective (talk) 18:47, 18 December 2025 (UTC)

•  Allow AI-media is good for FM if it is clearly labeled as such and realistically useful for the project. MZaplotnik^(talk) 19:03, 18 December 2025 (UTC)
•  Oppose I am not opposed to the use of AI as an assistive tool per se. AI can be useful for technical refinement, restoration or other clearly delimited support tasks applied to real, verifiable source material. However, I oppose the inclusion of purely AI-generated media at FMC at this stage. Current systems are examples of narrow AI, not artificial general intelligence. They do not possess understanding, intent or a concept of factual correctness; they generate statistically plausible output and are therefore capable of producing content that is convincing yet incorrect or misleading. For an encyclopedic project, this distinction is not trivial. Featured Media are implicitly trusted as high-quality, reliable reference material. Without a grounding in real, documentable sources, AI-generated media lack the epistemic accountability expected at this level. Even with disclosure, the fundamental problem remains: the content itself is not verifiable in the same way as human-created or source-based media. Until such systems can demonstrate reliable factual awareness and traceability - which current AI cannot - I believe that purely AI-generated media should not be eligible for Featured Media status. Best regards, -- Radomianin (talk) 19:55, 18 December 2025 (UTC)

• Because they do not possess understanding, intent or concept of factual correctness (and aren't designed for accuracy which is the main issue), one doesn't let AI generate things and just take the unmodified very first result and run with it.
  These are tools humans used and some people are pretty skilled in it with sophisticated workflows and good results. The human prompts the AI and selects the results and sometimes refines it further using the AI or other tools to make it good-quality and accurate (assuming accuracy is a factor in the illustrated topic).

Many people have this misunderstanding and/or make this false assumption. Prototyperspective (talk) 22:15, 18 December 2025 (UTC)

• Thank you for your thoughtful insights, Prototyperspective. I completely agree that AI is a powerful tool and that human guidance can greatly improve its results. For Featured Media, however, verifiability is essential. Narrow AI produces content purely statistically, without understanding, intent, or factual certainty, which makes it inherently difficult to confirm accuracy or trace its origin. Even careful prompting or post-processing cannot replace the accountability of human-created, documented sources. I therefore support a balanced approach: AI-assisted media can be used as supportive material, but for FM status, only content that is verifiably sourced and factually reliable should be accepted, while acknowledging that ensuring such verification is inherently challenging. This approach protects the integrity and trustworthiness of Featured Media. Best regards, -- Radomianin (talk) 22:45, 18 December 2025 (UTC)

Additional note: I have reflected further on my previous comment and decided to remove the struck section. I cannot accept any compromise: AI fundamentally undermines the principles Wikipedia and Wikimedia Commons are built on. I want to make it unequivocally clear that I am completely opposed to AI-generated media being included in any promoted content, such as Featured Media on FM or FP. Media that is AI-generated, even if it appears not to be created by AI, is not acceptable and must be strictly rejected. Best, -- Radomianin (talk) 10:02, 19 December 2025 (UTC)

•  Oppose per Radomianin, Cart, Abzeronow, and Moheen.   — 🇺🇦Jeff G. ツ please ping or talk to me🇺🇦 21:22, 18 December 2025 (UTC)
•  Oppose I'm sorry, I don't think AI has a place among our finest work. It has nothing to do with how accurate or verifiable it is, my objection is ethical. You can do really interesting things with AI, you can use it to analyze things, or make small funny clips, but an AI doesn't have a creative mind of its own. It has many very useful applications, but it can only put together what has been put into it, and what is put into an AI is the work of real uncredited artists. Along with a lot of other artists, I see that as theft. I don't want the finest work on Commons to be assembled from human artists through theft. When it comes to the work of humans we assess on FMC (and FPC), we are very picky and we don't allow people to take videos or images or sounds from others and call it their own. We've had content deleted from Commons for being stolen from other creators. But now users here are prepared to allow this just because it's been run through an AI mixer. Just because you can't easily recognize the parts from individual works by humans in an AI composition, doesn't mean it's not made up of that. AI media is "real" media, digitized, cut up into tiny bits of information and assembled into something that looks or sound new; but it's not created. --Cart (talk) 23:11, 18 December 2025 (UTC)

  Adding this link to illustrate: . --Cart (talk) 15:31, 10 January 2026 (UTC)

  I have to say, that's a very good argument. Ikan Kekek (talk) 23:18, 18 December 2025 (UTC)

  The prompter has the creative mind who envisions what they want, do ideation, apply prompting techniques, assembles the video, modifies results, selects results, adjusts the prompts, etc.
  And it is not theft to go to a museum or browse Google Images or watch TV and then learn from what you saw and produce things inspired by their styles, techniques, ideas, etc. Same for machines. It's very ethical to give the power to visualize concepts and ideas to the vast majority of common people who neither have chosen a career path in the creative industry nor made conventional art one of their main time-consuming hobbies. I consider it unethical to not do so or to discriminate against this production method. I do not think you made a good argument. Prototyperspective (talk) 00:26, 19 December 2025 (UTC)

 Support. I think AI is likely to end human civilization if global warming or nuclear war don't do so first. But that said, I think that AI media, as long as it's quite clearly described and not trying to pass for non-AI, should be eligible for nomination. I also think, though, that opposing anything AI just because it's AI is legitimate (as long as such opposition is not being used to outright delete files that are COM:INUSE, but that's not relevant here). In other words, to prospective nominators: go ahead and try, but expect to face a very high hurdle for actually passing. Ikan Kekek (talk) 23:12, 18 December 2025 (UTC)

• I feel like the opposers have made very good arguments, though, and that it should probably be necessary to somehow take ownership over an AI-assisted work in order to gain credit for it. it's definitely a problem to not have an author to credit for a potentially featured file. Ikan Kekek (talk) 23:32, 18 December 2025 (UTC)

•  Comment I think this is the cart before the horse. Let's see examples of AI generated content, and then have this discussion, rather than relying on knee-jerk reactions. Thanks. Mike Peel (talk) 23:23, 18 December 2025 (UTC)
  • I was surprised by how many people suddenly turned up here, but it looks like @W.carter: canvassed input, which is leading to immediate reactions creating unnecessary urgency. We have not been inundated by AI here, I'm not even sure if we've had any cases of AI content being submitted yet. Let's take the time to see how things progress. Thanks. Mike Peel (talk) 23:31, 18 December 2025 (UTC)

    Mike Peel, we had a nomination last year, and my wanting to know how the community felt about this was due to a question on a nomination we have right now. There are no guidelines for this, so it's best to ask. The reason I went to previous voter's talk pages and posted about this discussion, was that very few people have the FMC talk page on their watchlists (only 41 page watchers, as opposed to 608 for the FPC talk page, hard to get a discussion going with that few users knowing about it). FMC is unfortunately not as well visited as FPC. Please note that I deliberately waited to voice my own opinion until a number of other voters had already cast their votes. --Cart (talk) 23:49, 18 December 2025 (UTC)

    • The nomination last year went nowhere. The current nomination is clearly pointy and likely isn't going to get very far. There is no urgency here. Starting a discussion about the topic would have been the way to go, starting a vote is just causing people to vote one way or another based on their prejudices from elsewhere without listening to each other first. It's unnecessary drama. Thanks. Mike Peel (talk) 23:51, 18 December 2025 (UTC)

      This was not done for "drama", but simply to get people too come and discuss. Most of us know from experience that if you just invite users to discuss, almost no one will show up. Asking people to vote, they will show up though, because that is easier (the invite said "Share your thoughts and/or vote"). And, I have yet to see a vote on Commons that didn't turn into a discussion anyway. ;-) It was a way to get people engaged in the matter, and it seems to have worked. --Cart (talk) 23:59, 18 December 2025 (UTC)

      And now everyone has already expressed their support or oppose for this topic and others have to work to change their minds, instead of starting out with an open discussion. And meanwhile we don't actually have a functioning process here right now, see #No bot. Your opinion is the first !vote here, you didn't wait before sharing your thoughts, and whole approach to asking the question was not neutral. Mike Peel (talk) 00:09, 19 December 2025 (UTC)

      We actually do have a functioning process here since I've been doing the Bot's work manually for the last year and more. Check the logs. You don't need a Bot to keep a page running, we did that before the Bots were introduced, and when they malfunctioned. For more info see my comment on #book as Featured media?. I have been maintaining this page while we wait for some programmer to give us a working Bot. So far there have been no complaints about my work in keeping this project running. In fact participation here has increased since Eatcha left this in ruins and I took over maintenance. You are very malicious in questioning the work I put in here to keep this page running, like I have some secret agenda or something. Trust me, I do not! I have left maintaining FPC, due to more things to do outside Commons, but I hope I will be able to keep working on this page for a while yet. --Cart (talk) 00:24, 19 December 2025 (UTC)

      I'm striking out a sentence in my comment above as it was factually inaccurate, apologies for that, I made a mistake there. My bad. Mike Peel (talk) 18:45, 20 December 2025 (UTC)

    It's not canvassing if you contact FPC regulars irrespective of what opinions they might have, probably not even having much idea of what they will say. Ikan Kekek (talk) 00:03, 19 December 2025 (UTC)

    Exactly, I just went to the Logs and contacted people who had voted on FMC in the last 1-2 years. I didn't know if they had this page on their watchlists and no clue how they would vote. --Cart (talk) 00:09, 19 December 2025 (UTC)

    The way you set up the question and share it with others biases the outcome, and that's why discussion first is important. Mike Peel (talk) 00:12, 19 December 2025 (UTC)

    Since you know so well how this page should be run, perhaps you would like to take over my work here in maintaining this page? I would be happy to leave it to someone else. I thought I was doing voters here a favor to gain clarity in how to proceed on FMC in a new area of media. I did not expect to be attacked for my work. --Cart (talk) 00:26, 19 December 2025 (UTC)

    I strongly agree to your voting based discussion policy which I believe to have some fruitful outcome rather than just sharing thoughts.

    It really breaks my heart when someone's philanthropic hard work is questioned. I really hope you keep doing your good work for the benefit of people. Tisha Mukherjee (talk) 09:37, 19 December 2025 (UTC)

    I would say her message was very neutral, as RFC requests go. Mike, calling it canvassing is frankly ridiculous. JayCubby (talk) 01:41, 19 December 2025 (UTC)

• I honestly find it hard to follow Mike's accusation here. Cart's message reads to me as measured and neutral, and more like an attempt to gain clarity during a transitional phase than any form of canvassing. Regardless of differing views, it seems fair to acknowledge the substantial maintenance work Cart has been doing to keep this page running at all. -- Radomianin (talk) 06:52, 19 December 2025 (UTC)

• My central point here was that starting a discussion with a yes/no vote immediately polarises people, compared to a general discussion about the issue where people can talk through the consequences first. We need to seek a consensus viewpoint based on facts, which a yes/no vote makes a lot more difficult. Thanks. Mike Peel (talk) 18:51, 20 December 2025 (UTC)

  Your point of view is noted, since you have repeated it a couple of times here. However, so far you are the only one who seem to have this problem. But you are so vocal here, and with all the letters that the admin-bot is adding to your name you outrank us all, it looks like you are trying to make your opinion count for more than other voters. When the voting is done, I would like to add up votes pro and con and base the result on the highest number, but I suspect that you would like to do this differently. If so, it would be interesting to know that now. Perhaps you would like to call this whole thing as invalid and we should all go with what you think. --Cart (talk) 21:18, 20 December 2025 (UTC)

• The Wikimedia community makes decisions by consensus. Decisions aren't made based on how many votes each side has, nor the background of individual editors, they are based on the arguments presented and the community's values. I've expressed my disagreement in this thread about the framing of this conversation, and I've !voted below to express my view on this question. I trust that whoever closes this discussion will weigh my comments based on their merit, not on my background in the community. You should not be the one to close it since you asked the question, though. Thanks. Mike Peel (talk) 21:49, 20 December 2025 (UTC)

  Yes, sorry, I mis-wrote. I know I shouldn't close this. I only hope someone is interested enough to do the closing. --Cart (talk) 22:04, 20 December 2025 (UTC)

• So just out of curiosity: Since one argument here is that we shouldn't have this discussion when we've only had two AI nominations, when would the right time be? Should we have waited for the first 1-5 AIs to be promoted to FM? When People would be able to look back in confusion and point to the files already promoted and say that if they got promoted, why shouldn't we promote the rest. That's like waiting for accidents to happen before evaluations are made and preventive measures are taken. ("Let's not put up guardrails on this road through the ravine, we are convinced that people will drive safely and responsibly.") I personally think it's better to address an issue just before we see it's about to happen. --Cart (talk) 21:25, 19 December 2025 (UTC)

  Have such RfC been made for any other content type? Why are people so fearful of just letting people discuss and vote on things on a case by case basis the ordinary way.
  Would it be a catastrophe if a sophisticated well-engineered AI video was made FM? If so, is the reason for that because many people dislike these tools.
  Maybe some people who oppose this at this point based on what they've seen before didn't think of some other AI content type – let's say a data visualization video that has been made with AI.
  And maybe this production method even becomes the most widely-used method to produce such videos in 5 years, who knows now.
  There is no need to prohibit or explicitly allow AI content. It's not a good idea to always preemptively make broad polarizing exclusionary policies and you're describing things as if this was a safety policy to prevent accidents but nobody gets hurt if AI media gets FM. It's unlikely anyway for now there's any significant number of these files any time soon so it's a big fuss about nothing. Prototyperspective (talk) 21:45, 19 December 2025 (UTC)

  "a big fuss about nothing"? AI-created images is not like any other camera or editing technique we have encountered so far. The last time there was such a big shift in image making, was when cameras were invented and intruded onto the area of painting. That resulted in not a one great mix of techniques, but two rather distinct artforms, painting and photography. (Yes, I know there are cross-overs.) I'm not saying we should push fully AI-generated images out of Commons (and I agree with Giles that using some AI editing on photos is inevitable by now), I just don't think they have a place to be selected and voted on in FMC (or even FPC if this discussion ever gets there) where they are compared to and mixed up with human creations.

  When FMC was created, it was as a result of people at FPC feeling that this was a type of media so different from still images that it needed its own forum. Normal still photographers thought they couldn't evaluate such files sufficiently, or just didn't care to work with video or sound. It didn't start with a RfC, but as a suggestion with a short introduction followed by a vote. So that's the "other content type" you are looking for. Commons is a big place, and I for one wouldn't mind if the AI enthusiasts set up their own department on Commons, along with their own rules and featured voting system. A place where you could evaluate which AI systems were the best and exchange prompts with each other. That way the confusion would be minimized and AI contained within a circle of people that care for them and have their own view of what copyright and harvesting data is. --Cart (talk) 22:18, 19 December 2025 (UTC)

  I meant that - no AI media has yet been made FM - there are very few (less than 3 afaik) nominations for such and - it doesn't look either of these two things are going to change substantially any time soon. Fourth, any such media can be prevented from becoming FM via the usual procedure of voting in nominations where nothing on top is needed. Prototyperspective (talk) 22:29, 19 December 2025 (UTC)

  @Prototyperspective: I disagree. When the AIs and AI advocates start voting for them, it will be too late. We need to nip this in the bud, and preserve FM (as well as FP) as being guaranteed human-generated.   — 🇺🇦Jeff G. ツ please ping or talk to me🇺🇦 22:46, 19 December 2025 (UTC)

  Prototyperspective, so you do all this high-minded talk about the benefits of AI and how misunderstood it is, but when I offer an amicable solution that would enable you, and those who want to work with AI, to start something constructive for AI-generated images right now, you just go "noooo... I don't want to do anything". <<Heavy sigh!>> - said using an AI-generated Brit voice. I also agree completely with Jeff G. --Cart (talk) 23:37, 19 December 2025 (UTC)

  That idea is neither detailed, nor feasible, nor something especially clever to come up with that others haven't thought of before. And you just go "noooo... I don't want to do anything". <<Heavy sigh!>> is a bit over the top considering that I only didn't address this idea, would be interested if somebody would set it up, and did something similar on WikiBooks which got nothing but bad faith accusations on Commons so far and has no other editors yet, and already spend near 100% of my time on volunteering on Commons and open source/content projects. Are you asking for me to replace my sleep time?

  .

  And I'm not an AI enthusiast either, just in my view form nuanced opinion where I look not just for problems but also potential uses, principles to safeguard or apply, counterpoints to what people claim, and more. I don't care if people think I'm an AI enthusiast – I'm an open source technology enthusiast and roughly as annoyed by AI hype and LLM-misuse as I'm interested and engaged in making it benefit the projects.

  Don't feel offended just because I didn't address that idea or vent your frustration with little recognition of your work at FM out on me or whatever it is. If I should address your proposal: it's infeasible at least at this point and the forseeable future because way too few people submit way too little too low quality AI media on Commons and also because evaluating which AI systems are best is not something done on Commons and I linked a place that systematically compares models of a certain type and provided further reasons against a project of that sort at this proposal page for sth similar. And there won't be a FM equivalent just for AI, for example in all this time not even one such media was made FM and just 2(?) nominated and it wouldn't be much different there and featuring these there anyhow would have no effect whatsoever and would be meaningless. Another reason I didn't address it because it's tangential to the thread topic and just adds more wall of text. I was now trying to not respond anymore in this thread after having pointed out several false assumptions of voters and addressing a few claims. Prototyperspective (talk) 00:30, 20 December 2025 (UTC)

  Ok, thank you for at least taking time to respond in such detail. --Cart (talk) 00:59, 20 December 2025 (UTC)

•  Oppose per Radomianin. --August (talk) 23:35, 18 December 2025 (UTC)
•  Oppose per Cart. LLMs (they are not AI) are not creative. They do not frame with thought, they do not really address composition. They copy what already exists and mix it up in often bizarre ways that sometimes come out interesting. They are not what Feature Media is about. — Huntster (t @ c) 23:41, 18 December 2025 (UTC)

  Another comment based on false assumptions.

  1. generative AI that produce images, videos, and audio are not LLMs
  2. neither featured media nor good quality content requires creativity (and here the creativity would be in e.g. the ideation and prompt engineering etc of the prompter, not so much the AI) – e.g. various educational videos or static videos merely recording something interesting have been made FP and are not creative
  3. to "frame with thought" and "address composition" is not a requirement either and the prompter would achieve that and it's self-explaining that only those files where results are very good have any chance

  Prototyperspective (talk) 23:57, 18 December 2025 (UTC)

•  Strong oppose AI should be completely banned from Featured Media. Abzeronow (talk) 01:49, 19 December 2025 (UTC)
•  Strong oppose AI-generated media should be categorically excluded from FMC. FM represents Commons’ finest and most trustworthy work. Purely AI-generated media is not authored, documented, or grounded in verifiable reality; it is assembled from statistical recombinations of uncredited human works. This raises fundamental ethical and provenance concerns that disclosure alone cannot resolve. Commons does not accept derivative or misattributed human work at FMC, and the same standard must apply here. Allowing AI-generated media at this level would undermine both the ethical integrity and curatorial credibility of Featured Media. --Moheen ^{(keep talking)} 04:36, 19 December 2025 (UTC)
•  Oppose per Radomianin & Cart. -- Yellow|ø   11:09, 19 December 2025 (UTC)
•  Oppose Responsible use of AI is not yet normalized, good practices have not yet been established. As of now, I am voting against because the risk of spreading low-quality, misleading content is too great. This may be reconsidered in the future.Sillerkiil (talk) 12:39, 19 December 2025 (UTC)

  Low-quality, misleading content would simply be voted against in the FM nomination.
  Like Mike Peel, I can't see why this needs a RfC to exclude this production method preemptively instead of simply letting users vote against the nomination (of which there afaik was only 1 so far, underlining how unneeded this is) or creating an RfC if this actually becomes an issue. I don't think there is any other tool or production method that has been prohibited/excluded so far and the nomination discussions are the place to reject them. Prototyperspective (talk) 16:44, 19 December 2025 (UTC)

•  Oppose 100% AI generated content. Neutral for normal non AI content slightly "enhanced" with AI (such as noise reduction in video, etc.) -- Giles Laurent (talk) 16:53, 19 December 2025 (UTC)
•  Oppose As per Radomianin, Cart, Abzeronow, Moheen and JeffG--Don (talk) 19:04, 19 December 2025 (UTC)
•  Strong oppose for current use of Gen-AI, especially per Moheen. The ethical concerns are too great with current use. Maybe the medium will evolve in the future, photography was poorly regarded as an art medium at its beginning. But as it stands, the provenance concerns are too great and I don't think it should be a door that is opened with the current IP theft that allow the existence of Gen-AI, especially when training sources aren't readily available. --Alexis Lours (talk) 22:03, 19 December 2025 (UTC)
•  Allow If we have to cast a !vote now (per my comments above I'd have preferred to see a discussion first), then my !vote is to see what happens first, and that defaults to being more permissive and letting AI generated candidates stand and have fair discussion about whether they are Featured quality or not, rather than them being excluded automatically. I'm not a fan of AI, and I also have a lot of concerns about the ethics behind it and the environmental impacts that it is having. However, the same is true for other contributions - when we upload our own media here, we don't credit the developers of the cameras we use, or the techniques we've learnt from others, or the feedback we've heard from others about how to post-process them, or the environmental consequences of our travel to capture that media. Let's see what ends up here first, and ask critical questions then - and after we've seen how things are working out, then we can make a decision based on evidence rather than emotions. Thanks. Mike Peel (talk) 19:08, 20 December 2025 (UTC)
•  Oppose I was going to come in neutral on the basis of "can't imagine anyone actually believing that slop is among the best media on Commons, so probably no need for a prohibition" but now I see it's actually getting some support. If Wikipedia had its own FMC process, where illustrating articles is of paramount importance, I could see some support for notable AI-generated stuff, but here we're aiming to highlight the best Commons has to offer, where educational value is just one component alongside technical quality and "wow factor". Not opposed to revisiting this down the road, but for the time being I just have a hard time understanding how we'd evaluate technical quality and wow factor for something AI-generated, and difficulty seeing how we could identify it as among the best files Commons has to offer. — Rhododendrites ^talk |  20:48, 20 December 2025 (UTC)
•  Oppose per Moheen and JeffG – “We need to nip this in the bud, and preserve FM (as well as FP) as being guaranteed human-generated.” Neutral for purely auxiliary use of AI tools for post-production etc., e.g. noise reduction in videos. – Aristeas (talk) 20:33, 21 December 2025 (UTC)
•  Oppose As per Radomianin, Cart, Abzeronow, Moheen and JeffG. AI-generated media is typically the random result of semi-random processes. And also this is bad downshift from our real life into the Matrix or into the Terminator Worlds. -- George Chernilevsky ^talk 20:43, 21 December 2025 (UTC)
•  Neutral Judge works by their quality, not by their fabriation method. --Trade (talk) 00:34, 22 December 2025 (UTC)
•  Strong oppose as per Radomianin's additional note, JeffG, Cart, and Moheen, with the caveat mentioned by Giles Laurent. I would like to disclaim that, though I am not active on FMC (I am active on FPC, and if this affects the validity of my vote I will not contest it), I believe that allowing this precedent to be set here would affect other pages and projects, as well as much future discussion on the platform, which is why I feel compelled to vote after considering these possibilities. I agree entirely with JeffG's reply to Prototyperspective, which was also reiterated by Aristeas. – Julian Lupyan (talk) 20:23, 22 December 2025 (UTC)
•  Support On the condition that the candidates should have a Wikipedia article of their own, and their quality or impact is relevant. I can only think of a handful of examples complying these conditions. --Bedivere (talk) 00:16, 26 December 2025 (UTC)
•  Oppose AI-generated media is not worth spending human time to vote on. --Lupe (talk) 21:06, 27 December 2025 (UTC)
•  Oppose — Preceding unsigned comment added by Ooligan (talk • contribs) 19:46, 28 December 2025 (UTC)
•  Strong oppose Per Radomianin, Cart and Moheen. Additionally, featured media should constitute deliberate, well-thought-out, high-quality work. Generative AI bypasses this quality and effort standard for a subpar alternative, which fails to align with the goals of featured media. It's moon (talk) 23:42, 30 December 2025 (UTC)
•  Neutral For the time being, I support keeping the current status quo i.e., not explicitly allowing or disallowing AI-generated media in the rules, and having a case-by-case discussion in the nomination:

• The discussion above is fascinating and highlights key issues with AI (e.g., ethics, traceability, AI-generated vs. AI-assisted). But it also showcases several misconceptions about how AI works, what it is currently able to do, and what it will likely be able to do in the very near future. Continuing the discussion over a few more concrete examples would therefore be useful
• FM (and FP) are part of Wiki Commons and should not deviate from it on key policy matters. If Commons turns into a human-only repository, then FM and FP should mirror that. Until then we should continue to showcase the best media on Commons, irrespectively of how it was generated
• The recent AI nomination is clearly marked as AI-generated in both the description and the categories: should we add a disclosure requirement to the rules to ensure future nominations follow the same path? In my opinion, this disclosure should cover all forms of media manipulation, both for AI and non-AI tools

--Julesvernex2 (talk) 10:11, 31 December 2025 (UTC)

@Julesvernex2 Thanks for your comment, to be on the same page I'd be interested in learning what misconceptions you believe to be present in the discussion above. It's moon (talk) 12:32, 31 December 2025 (UTC)

Hi It's moon, I think they fit into two big buckets (both of which almost always come up in discussions about AI, and are a natural reaction to a technology that is evolving at an unprecedented pace):

• AI produces only low-quality content: current algorithms can produce content that is indistinguishable from that of humans (for text), getting there as we speak (for images and sound), or likely to get there in the very near future (for video). As for any other tool, AI is also perfectly capable of producing slop if used improperly or carelessly. Going from slop to accurate human-level output is easier than it seems, as we can simply ask the AI itself for suggestions on the prompt engineering and iteration techniques that best fit our intended output. Don't take my word for it, give it a go!
• AI cannot produce new content: it is true that AI algorithms are trained on existing human-generated content and generate outputs by statistical inference. However, that doesn't directly imply they're limited to recombining existing content, much like a child is inspired but not bound by what he or she can observe. Our brains and AI work in radically different ways, but can we claim that only the former has creativity and intent if both produce similar content? Ceci n'est pas une pipe?

I hope these points are useful but, for the discussion at hand, I don't think they offset the above mentioned problems with ethics and traceability. However, when and if we change the rules, we should remind ourselves that shutting the door to disruptive technologies is rarely a good idea. Wikipedia and other reference websites have already taken a hit from Google's AI-generated summaries on top of search results, and Wiki Commons will likely go the same route when generating an image becomes easier than searching for one. Julesvernex2 (talk) 15:06, 31 December 2025 (UTC)

Thanks for a substantive, thoughtful post. I hope you don't mind if I address just one point:

"Our brains and AI work in radically different ways, but can we claim that only the former has creativity and intent if both produce similar content?"

Yes, of course, because AI does not think or create, it just imitates and quotes and sometimes synthesizes. Ikan Kekek (talk) 18:39, 2 January 2026 (UTC)

Well, in a sense so do we, as humans do not create things from ex nihilo. Taking a couple of examples that I think are dear to both of us, a photograph is a reinterpretation of an existing scene, and music is creativity bound by a set of rules. AI works within the same constraints, although following a completely different process that is not well understood and perhaps will never be. My argument above is that the process doesn’t matter, what matters is the outcome. Using science as an example, where it’s easier to identify novelty, there are plenty of examples of AI creating something that did not exist: protein folding; new enzymes and chemical compounds; out-of-the-box Go strategies. In an example that hits close to home, a few months ago I read about a PhD student that witnessed an AI algorithm replicating his entire thesis (a new way of estimating the mass of a black hole, if memory serves me well) in just a few hours… Julesvernex2 (talk) 20:11, 2 January 2026 (UTC)

In my opinion, if we are using the argument that the outcome is all that matters, the desired outcome for media on this platform is that it is an accurate, educational representation of whichever subject matter it depicts. In terms of featured media, the media has to be the best of the best in terms of accuracy and educational utility, and both of these metrics seem not to be met by AI content. Even if the process is disregarded, I believe the outcome is still contrary to what is sought for by those who use Wikimedia Commons. – Julian Lupyan (talk) 21:37, 2 January 2026 (UTC)

I agree that those two objectives are not yet consistently met by AI. But I think what follows is that we don’t need to change the rules to explicitly ban the use of AI, we just need to continue to reject AI nominations until those goals are met. Personally, as a human that uploads content to Commons, I’m in no hurry to see that happen (less competition!) but I think end-users will appreciate the broader offering. Maybe that’s the niche Commons needs to secure its existence: a repository which validates its content for accuracy and educational value, no matter how it was generated? Julesvernex2 (talk) 22:25, 2 January 2026 (UTC)

Only if you make these two false assumptions:

1. that are only factual subjects but not things like say depictions of fantasy objects
2. that there is no prompter who uses the AI to get the desired outcome by for example selecting from the results, changing the prompt, and/or editing.

A third false implicit assumption is that people are unable to take what you said into account in the nomination discussion which is the place to evaluate the media and discuss accuracy and utility issues. That third false assumption can get bolstered by unsubstantiated fear-mongering of Commons deliberations – like FM discussions – in the future maybe getting undermined/flooded by bots (plus humans whose points some people like to dismiss by dehumanizing/devaluing them as mere "AI advocates"). Nothing of that sort has happened and nothing indicates it will…and if it does, one could still have this RfC at that point but it's not much more than sth similar to unfounded hysteria imo. A fourth issue is people thinking their notions and narrow personal experience of what AI media is, applies to all AI media, including in the future, instead of there maybe being sorts of AI media they haven't even thought of could exist. Prototyperspective (talk) 23:06, 2 January 2026 (UTC)

Wikimedia Commons is not a generic file host, it is the repository of an encyclopedia, among other projects where accuracy is important. Generative AI works by plausibility, not factuality. It generates what looks or seems right, not necessarily what is right. Even with a prompter willing to improve the prompt, that doesn't guarantee the output is free from hallucinations and bias, which are often hard to spot.

Additionally, I can't stress this enough, by generating and publishing AI generated fictional content or art, especially if it's high quality and especially if you release it to public domain, you aren't just creating media effortlessly using a dataset of artists' hard work which they never consented to, but you are also competing against them and effectively devaluing their work. It's moon (talk) 02:52, 3 January 2026 (UTC)

I agree. As mentioned by Julesvernex2, accuracy and educational value are two metrics that are not currently well fulfilled by AI. I agree with him that, ideally, it would be best not to outright prohibit AI media entirely, but for people to decide whether it meets this criteria. However, seeing as for now, and for the near future, AI media is not meeting this criteria, I feel as though it is justified to add a restriction to the rules, as it will save unnecessary case-by-case review for media that is inherently (for the time being) not featurable. Functionally, if we continue rejecting AI nominations until these goals are met, it would be the same as if we restricted them, and altered these restrictions once said goals are met. I prefer the second option, because it is more concrete, and will save time that could be used to review more useful media, without having to restate consensus and contingencies each time AI media is nominated. – Julian Lupyan (talk) 03:03, 3 January 2026 (UTC)

A small but relevant caveat on my position: I said that these two metrics are not yet consistently met. By not adding an explicit ban on AI, we can vote favorably on AI-generated content that already fulfills the criteria, while rejecting others that don’t. Judging the accuracy and educational value of AI-generated content isn’t harder than doing so for media that has been manipulated with non-AI tools that are also not banned (e.g Photoshop, After Effects, Audition) Julesvernex2 (talk) 08:04, 3 January 2026 (UTC)

It's moon, I think those ethical concerns are a valid reason for one to individually oppose an AI-generated nomination, but not to change the rules to ban AI before and if Commons does the same thing. FM is part of Commons and benefits from having its featured media prominently displayed on the Commons main page, so it shouldn’t deviate substantially on policy matters Julesvernex2 (talk) 09:22, 3 January 2026 (UTC)

I think my ethical and non-ethical concerns are valid to oppose AI being featured on Commons / be elegible to compete against other media. I am not the only one opposed as well. I disagree with the premise that FM must mirror general Commons policy. The entire purpose of Featured Media is to set a higher standard than the general repository. It's moon (talk) 13:45, 3 January 2026 (UTC)

Your comment does not relate to mine (to which you apparently replied). None of what you said addresses or takes into consideration anything I said or implied and I'm well aware "Generative AI works by plausibility, not factuality … It generates what looks or seems right, not necessarily what is right" – apparently you did not read or understand the two false assumptions which addressed exactly that.
And the invention of the photo camera didn't devalue art and that it may have devalued some particular types of art such as realistic portrait painting and upvalued&created other kinds of art (all of which is not an issue and no reason to discriminate against one particular production method). Moreover, not all media where AI is used in the workflow is created "effortlessly" so this false claim further reveals your bias.
Lastly, re "The entire purpose of Featured Media is to set a higher standard than the general repository" then vote on nominations accordingly. Prototyperspective (talk) 09:51, 4 January 2026 (UTC)