User talk:Athanelar/Identifying AI-generated text#rfc 08DBF24

Since this is your proposed essay, and I don't want to potentially change the meaning without your input, I wanted to ask you if you could add something to the aforementioned section. It reads like anyone who uses AI understands the drawbacks of AI and all of the relivent P&Gs. Most new editors wouldn't know any of this. So, could you add that people should explain to the AI using editor what they did wrong/direct them towards some of our AI and AI relivent P&Gs. I don't want any newcomers bitten. Mikeycdiamond (talk) 19:07, 26 November 2025 (UTC)[reply]

The purpose of this guideline isn't to state that generating articles with AI isn't allowed, nor is it to explain why using AI for other purposes is discouraged, or the problems that AI usage can create on Wikipedia. That's what WP:NEWLLM and WP:LLM are for.

The scope of this proposal is purely to serve as a supplemental to NEWLLM. The idea is that NEWLLM says AI-generated articles aren't allowed, and this essay would serve as an explanation of how we classify an article as 'AI-generated' in order to remove it based on NEWLLM.

Basically the ecosystem would be; if you want to link someone to something that explains why AI is problematic on Wikipedia, send them to WP:LLM. If you want to nominate an article for deletion because it's AI-generated, you reference WP:NEWLLM as the guideline which is being violated. If somebody asks 'but what's the standard the community uses to judge that an article is AI-generated,' you'd send them here, and specifically reference which section you're using to evidence the article being AI generated. I.e., in an AfD discussion you might leave a comment like "Delete. Article text appears to be primarily AI-generated in violation of WP:NEWLLM. The article seems to be AI generated because the text (shows [x] number of WP:AISIGNS/contains communication directed at the user/etc) which demonstrates that it's probably AI as per [this essay]" Athanelar (talk) 20:54, 26 November 2025 (UTC)[reply]

I think you might have misunderstood me. I am not asking you to put an explanation into the proposed essay, just a recommendation that someone explains/links the relevant P&Gs before assuming the AI was used in bad faith. Mikeycdiamond (talk) 21:48, 26 November 2025 (UTC)[reply]

I've tweaked the last section somewhat to add this information. Athanelar (talk) 22:20, 26 November 2025 (UTC)[reply]

You may want to pull the ineffective indicators section from WP:AISIGNS, or include a section of things that point away from AI use. The only one I can think of at 2am though is that they don't do parenthetical asides, as they don't have side thoughts. ~ Argenti Aertheri^(Chat?) 06:54, 27 November 2025 (UTC)[reply]

Please consider joining the feedback request service. An editor has requested comments from other editors for this discussion. This page has been added to the following list: Wikipedia policies and guidelines When discussion has ended, remove this tag and it will be removed from the list. If this page is on additional lists, they will be noted below.

This has been up for a bit over 24 hours and has had a fair few eyes on it with no strong objections and some good tweaks, so I'm moving forward to RfC.

The RfC closer for WP:NEWLLM stated that a community consensus on identifying AI-generated text would be necessary in order for that guideline to be properly enforced.

The question for this RfC is thus: should this proposal be accepted as a supplementary essay to WP:NEWLLM (and future AI-restricting guidelines) to serve as a consensus standard for how to identify AI-generated text? Athanelar (talk) 00:19, 28 November 2025 (UTC)[reply]

• No, this is too soon and possibly redundant to WP:AISIGNS. I would also like to point out that the WikiProject AI Cleanup project page states The purpose of this project is not to restrict or ban the use of AI in articles, but to verify that its output is acceptable and constructive, and to fix or remove it otherwise. SuperPianoMan9167 (talk) 00:30, 28 November 2025 (UTC)[reply]
• Support While there is overlap with AISIGNS, this essay also focuses on AI detectors and what to do if you suspect AI is being used. Mikeycdiamond (talk) 00:39, 28 November 2025 (UTC)[reply]
• No per SuperPianoMan9167. It's far too soon to think about promoting this - it needs to spend a good deal of time (on the order of months not hours) as a normal essay getting tweaked as it encounters the real world and more editors become aware of it. I sort-of understand the desire to rush things with regards AI-policies and guidelines but it is far more important than we get things right than we do things quickly. Thryduulf (talk) 09:39, 28 November 2025 (UTC)[reply]

  I'm concerned about the idea of NEWLLM floating around for 'months' in a state of uncertainty; and if people think this is a good base then it's something that can be tweaked and expanded on after acceptance anyway. The main guideline is already in force and is evidently lacking some in-demand supporting infrastructure, I think it makes sense to prioritise patching the hole quickly and work from there. Athanelar (talk) 09:46, 28 November 2025 (UTC)[reply]

  It wouldn't be "floating around in a state of uncertainty" it would be a normal essay. Essays can be tweaked far more easily than supplements, because changes to the latter need to be certain that they continue to represent the consensus that was agreed when they were promoted. The lack of supporting infrastructure is a problem caused by promoting a poorly thought-through vague proposal before it was ready, you don't solve that by supplementing it with with something that is equally unready. Thryduulf (talk) 09:55, 28 November 2025 (UTC)[reply]

  It wouldn't be "floating around in a state of uncertainty" it would be a normal essay I mean the WP:NEWLLM guideline, not this essay. Athanelar (talk) 10:39, 28 November 2025 (UTC)[reply]

  As I said, that's the unavoidable consequence of promoting something vague without fully thinking it through first. It cannot be fixed by rushing in supporting infrastructure that is similarly unready. It can only solved by either repealing the guideline, dealing with it for now and then replacing it with a guideline that is fully thought through, workshopped and developed, or dealing with it for now and adding the supporting infrastructure after fully thinking it through, developing and workshopping it. Thryduulf (talk) 11:10, 28 November 2025 (UTC)[reply]

  +1 ~ ToBeFree (talk) 20:40, 30 November 2025 (UTC)[reply]

• Not yet - I support the idea in theory, and I understand why this exists, but I think the essay needs some more workshopping. Right now it is sort of a mix of summaries of our other pages about identifying AI text, a brief section about how to interact with editors using AI that doesn't address the very common case of AI text added by no-longer-active editors, and not much about gathering consensus about a particular piece of text which was what the RfC close mentioned. Gnomingstuff (talk) 19:45, 30 November 2025 (UTC)[reply]
• No - needs much more time. I'm also not sure how this doesn't overlap with Wikipedia:AISIGNS. User:Bluethricecreamman (Talk·Contribs) 05:36, 1 December 2025 (UTC)[reply]
• Not yet, it's definitely a good start, but it needs more time to ferment before becoming a supplement. ~ Argenti Aertheri^(Chat?) 05:38, 1 December 2025 (UTC)[reply]