Wikiwand AI

Talk:Mexico City

From Wikipedia, the free encyclopedia

More information WikiProject Geography To-do list: ...
Close

Possible unreviewed AI use

Hi -- I have tagged this article as potentially AI generated, specifically edits by user Synthetic.Cognizance. They display many tells of AI writing, and a sockpuppet investigation seems to be further evidence. Needs review, fact-checking, tone check, etc. Gnomingstuff (talk) 18:34, 13 August 2025 (UTC)

@Gnomingstuff: The notices on top of the article tend to be seen as a major problem throughout the article. In fact, when you add a template that says "This article may incorporate text from a large language model. Such text potentially includes hallucinated information or fictitious references" implies that it, indeed, lacks creditbility. You are saying that because one person, exclusively, who just added one picture and one microparagraph to this article guarants the disclaimer the readers must interpret as "warning, some text (which we won't reveal) might be LLM". This template is getting out of control, to be honest. (CC) Tbhotch™ 18:59, 13 August 2025 (UTC)
That is exactly what I am implying. The AI use is what's gotten out of control, and it would still be there template or not; editors and readers may as well know about it. At any rate, if the edits are so small-scale then fixing them should be a very quick fix and then you can remove the tag. You could probably have fixed it in the time it took to write this comment.
(To address the obvious question: Right now, my focus is on identifying undetected AI edits so that they can be in one centralized place. I've taken more direct action when it involves unsourced material on WP:BLP articles -- i.e., when the guideline is to just revert, no questions asked -- but the most pressing issue right now is actually finding the stuff.) Gnomingstuff (talk) 19:09, 13 August 2025 (UTC)
  • "if the edits are so small-scale then fixing them should be a very quick fix and then you can remove the tag." Something that you, in fact, had to to first.
  • "You could probably have fixed it in the time it took to write this comment." Yes, in the time you wrote this answer, I fixed it in three other articles, which, once again, you should have fixed yourself instead of tag-bombing dozens of articles.
(CC) Tbhotch™ 19:24, 13 August 2025 (UTC)
Thank you for fixing those issues.
That being said, given the enormous amount of undetected LLM edits that exist, there are basically only three options: ignoring it, tagging, or mass deletion. Reversion is usually not an option since this stuff tends to get buried deep in edit history and tweaked without addressing the root cause. Fixing them one by one means the rest just go undetected that much longer. Of those options, mass deletion is generally considered much more disruptive -- especially when there's no actual proof of anything -- and not fully supported by policy except on BLP articles. (WP:CHATGPT is an essay.)
As far as tagbombing, here's what the text of WP:TAGBOMB -- which is also an essay and not policy -- actually says: "Adding multiple tags without explaining the reason is disruptive." (Bolding mine.) See also WP:OVERTAG: "However, it is not actually "drive-by tagging" so long as [editors] explain the rationale for the tag on the talk page." So, when the potential LLM use is not immediately obvious or easy to find -- basically, when I had to go through 5 pages of stuff to find it -- I've been explaining the rationale and mentioning the involved user and/or linking to the original diff so others do not also have to go through 5 pages of stuff.
One thing I should be doing more of is removing redundant tags -- for instance a lot of people add the "reads like an essay" tag when the reason something reads like an essay is because it's ChatGPT. When I get a chance I'll go through my recent edits and do that. Gnomingstuff (talk) 19:59, 13 August 2025 (UTC)

Related Articles

Timelines

Top Qs

Fact Checks