Wikipedia:Miscellany for deletion/Wikipedia:Case against LLM-generated articles: Difference between revisions

Wikipedia:Case against LLM-generated articles (edit | talk | history | links | watch | logs)​

In December 2023, an editor created Wikipedia:Why you shouldn’t write articles with ChatGPT, according to ChatGPT. The concept of this essay is for it to be an LLM-outputted text about why not to post LLM outputs on Wikipedia.

In October 2025, another editor created Wikipedia:Case against LLM-generated articles (the page nominated in this MfD), also an LLM-outputted “essay” about the same thing. The editor did so apparently unaware that the concept had already been used.

So now we have two AI-generated essays with the (frankly juvenile) “I made the AI write about how bad it is” punchline.

The reason for this MfD nomination is that the first essay, from 2023, is good for what it is, and the second one is not good and is redundant and confusing.

It is confusing because it is a partly unattributed paraphrase of the real human-written essay Wikipedia:Large language models. In two places it explicitly quotes it. It is not legitimately novel AI-generated text but is very closely derived from an existing human-written text. So the concept is muddled, because this simply labels a paraphrase of an existing serious essay as humorous, which it is not intended to be, and the paraphrase conveys the same message and is not funny in any way. (The 2023 essay at least has “ongoing improveability” as an example of a hallucinated word.) And if someone believes that there is something funny about the content of the 2025 essay because of the humor tag, that person would only be misled.

This appears to have happened because LLM chatbots have gained the ability to perform web searches and read web pages in the meantime, and the essay “Wikipedia:Large language models” has been cited in various places. This gives chatbots the ability to closely paraphrase it. Whereas the original 2023 essay with this concept is “authentic” AI slop, this 2025 attempt is a more competent attempt to generate a text on this topic by closely sticking to something created by humans, enabled by these new capabilities which did not exist in 2023. —Alalch E. 03:02, 11 October 2025 (UTC)[reply]

We have plenty of overlapping humor pages. Deleting it sounds too excessive when we could simply merge it or keep it. Redundancy and confusion aren’t exactly fatal flaws when it comes to an essay. Gommeh 📖   🎮 04:43, 11 October 2025 (UTC)[reply]

It can’t be merged into anything because the “essay”‘s concept is to retain AI generated content, not alter it by stringing it together with something else. Deleting is not excessive because the content is AI slop not fit for human consumption, it should not be read by anyone, it brings no benefit to the project and may only confuse, so it is a net negative. —Alalch E. 10:18, 11 October 2025 (UTC)[reply]

So… your only reason for deleting this is because it is “AI slop”? Gommeh 📖   🎮 13:49, 11 October 2025 (UTC)[reply]

Please don’t do that. I wrote cca 400 words explaining the reason to delete and you selected two words from what I wrote as if it represents everything I wrote. That is quoting out of context. Refer to the nomination and my reply to your comment above for the reason to delete. Better yet, tag this bad page that you have created with G7. —Alalch E. 14:01, 11 October 2025 (UTC)[reply]

I understand your point; the problem is, those are not valid reasons to delete the page. You’re arguing that it should be deleted because you don’t think it’s funny (or something like that). Humor is subjective. There are lots of things marked as humorous essays on here that I don’t find funny at all, but that doesn’t mean they should be deleted. Gommeh 📖   🎮 19:42, 11 October 2025 (UTC)[reply]

I appreciate that you have stated that you understand my point, which means that you are interested in understanding my point, but the way to understand my point is to read what I wrote. While you claim that you understand my point, you do not in fact understand my point. But I can help with that: Read what I wrote as being the point, and no special exegesis is needed beyond reading, because what I wrote is immediately understandable. What I wrote is the point and nothing else is the point. By reading what I wrote, you will immediately understand it. And after reading and understanding, you will no longer feel the need to deform my point into “it is ‘AI slop'” and “don’t think it’s funny”. Instead, you will know that my point is that the page you have created is redundant and confusing. You will further know that my point is that it is confusing because it is a partly unattributed paraphrase of the real human-written essay Wikipedia:Large language models. In two places it explicitly quotes it. It is not legitimately novel AI-generated text but is very closely derived from an existing human-written text. So the concept of an essay as an LLM-outputted text about why not to post LLM outputs on Wikipedia is muddled because this simply labels a paraphrase of an existing serious essay as humorous, which it is not intended to be, and the paraphrase conveys the same message and is not funny in any way. (The 2023 essay at least has “ongoing improveability” as an example of a hallucinated word.) And if someone believes that there is something funny about the content of the 2025 essay because of the humor tag, that person would only be misled. You will also know that my point is that the essay can’t be merged into anything because the “essay”‘s concept is to retain AI-generated content, not alter it by stringing it together with something else. Deleting is not excessive because the content is AI slop not fit for human consumption; it should not be read by anyone. It brings no benefit to the project and may only confuse, so it is a net negative.Now that you have got a new chance to understand the point, you have gained a new chance to tag the page for G7 deletion and undo this bad action of yours in creating this bad page.—Alalch E. 21:42, 11 October 2025 (UTC)[reply]

No, I completely get the point you are trying to make. The trouble is, none of what you’re saying is a valid reason for deletion under WP:DEL-REASON. The fact that you also stated that this essay should be read by no one is WP:IDONTLIKEIT as well, which is an argument to avoid. Additionally, as I said before humor is subjective. Granted WP:ITSFUNNY is primarily targeted towards pages in article space, but the main point it makes is that things shouldn’t be deleted based on humor (or lack thereof) alone. Gommeh 📖   🎮 00:30, 12 October 2025 (UTC)[reply]

I am not trying to make a point, I have made a point, and I loathe your attempts to mischaracterize my arguments through quote mining and strawman. The trouble is your creating this trash and your tendentiousness in this discussion and lack of accountability. WP:PRJDEL: Editing cannot improve this page because its concept is that it was generated by an AI chatbot and so editing it only undermines its concept and weakens and muddles it further. A projectspace page that is bad and does not serve the community and cannot be improved through editing should be deleted. At the very least this “essay” should be moved to your user space. —Alalch E. 00:38, 12 October 2025 (UTC)[reply]

The point is that it doesn’t need to be unedited, only that it needs to look like AI. It can be improved without deletion. It is supposed to sound verbose and overconfident like an AI to drive home the satire. Gommeh 📖   🎮 02:20, 12 October 2025 (UTC)[reply]

No, that concept does not work because the page is purportedly funny only because it is LLM-originated and speaks against LLMs, and the reader does not know that humans might have edited some portions to emulate the mannerisms of LLMs. Irrelevant. The only thing the reader is facing is a series of serious statements about the risks of AI use on Wikipedia generally taken directly from an actual, human-written, essay with a lot of community support presented as funny. That is misleading. Projectspace pages cannot be misleading. No type of editing can fix this page you have created.

Actually, the only way to fix it would be to remove the humorous tag and remove the cringey adulterations of the original LLM output such as your addition of inane fake fictitious references to make the actually competent output of the LLM worse than it actually is, further misleading users by misrepresenting how LLMs function and their capabilities. One such adulteration you added is for example this reference: “Human, Real (July 12, 2024). Definitely Real Book That I Didn’t Just Make Up. pp. 45–234.”

That is terrible. Extremely unhelpful page. Do you finally understand? —Alalch E. 02:51, 12 October 2025 (UTC)[reply]

Keep. These are not reasons to delete. There are ways to organise similar essays without deletions. SmokeyJoe (talk) 08:36, 11 October 2025 (UTC)[reply]

Comment – there may have been a typo there. Can you elaborate on what you mean, @SmokeyJoe? Gommeh 📖   🎮 19:42, 11 October 2025 (UTC)[reply]

There is a long occasional history of editors complaining about too many essays. The answer is to organise them, even merge and redirect. Rarely is deletion of an essay a good idea. Also, challenging MfD to come up with judgements in 7 days is not a good use of MfD, and I retreat to: “The reasons offered for deletion are not valid”. They are just criticisms.

I am also unhappy that the obvious middle ground is unmentioned, “Userfy”. SmokeyJoe (talk) 00:42, 12 October 2025 (UTC)[reply]

Redirect to Wikipedia:Why you shouldn’t write articles with ChatGPT, according to ChatGPT 88.97.192.42 (talk) 22:36, 11 October 2025 (UTC)[reply]

Just realized I forgot to !vote. Keep per SmokeyJoe. Gommeh 📖   🎮 23:09, 13 October 2025 (UTC)[reply]