Talk:Anaptychia ciliaris

Anaptychia ciliaris

Article (edit | visual edit | history) · Article talk (edit | history) · Watch • Watch article reassessment page • Most recent review

Result pending

Per this discussion, this article was found to have been, at least in part, LLM-generated. Wikipedia cannot in good conscience reward LLM-generated articles with GA status. Bgsu98 (Talk) 02:12, 24 April 2026 (UTC)

Retain Since no actual issues with the article were noted in the above nomination, seems like a procedural retain close is appropriate. Silverseren^C 02:30, 24 April 2026 (UTC)

As the original GAN reviewer, I vote for delist. I obviously cannot do a proper review when I do not know that the content is LLM-generated. LLMs make unexpected mistakes a human writer would never make, and a much more rigorous source review would be required (that might require even more time than was invested in generating the article in the first place). I had a look at the first paragraph, and checked this part:

It was formally described as a new species in 1753 by Carl Linnaeus, as Lichen ciliaris. Linnaeus described it as a somewhat erect, leafy, grey lichen with linear, fringe-like segments that are ciliate (having hair-like structures, or cilia). He said it resembled a tree moss with hairy edges and small shield-like structures (apothecia). Linnaeus cited multiple references that described the lichen similarly, emphasizing its larger size, hairy characteristics, and shielded appearance. Linnaeus noted that this lichen is found on trees in Europe.[4]

I don't see that this is covered by the source. Linnaeus does not seem to describe it this way himself as stated; all of this is based on his quotes of other works. Such (very significant) issues are typical for any LLM-generated article. I also do not think that we have to list "actual issues" to denominate an LLM-generated article. These issues will be found if we look for it, but that would take a lot of time, and I prefer to invest that time to help actual human editors. --Jens Lallensack (talk) 05:55, 24 April 2026 (UTC)

Jens Lallensack: 1) You have no evidence of any of it even being LLM-generated, nor do you know how Esculenta writes their articles. 2) The source you brought up is literally one that an LLM wouldn't be able to even read or doing anything with. So it even more clearly showcases that there isn't evidence of LLM being used to write it. 3) The source is literally Linnaeus' reference book for describing plant species, Species Plantarum, and, yes, it uses other sources to create a conglomerate description. That's how Linnaeus made the work. I don't see how that makes anything that you quoted inaccurately. It is him choosing that description for the species. Silverseren^C 22:23, 24 April 2026 (UTC)

@Silver seren nor do you know how Esculenta writes their articles: actually, yes we do, and this diff was included in the first post in the WT:GAN thread Bgsu98 linked to in their GAR rationale. Also see for further relevant evidence of LLM generation; Bgsu98 has only GAR'd articles among those 32 GAs with highest likelihood of LLM-generated text. NicheSports (talk) 00:31, 25 April 2026 (UTC)

I know full well she had a personally made AI generation system that she used and reviewed for all articles she made. And yet every attempt I see to bring up fault in these GAs thus far has just made the claimant look like they don't understand the references they're reading. Above is yet another example of that. Unless you have actual examples of issues with the article, then you don't have an argument. Silverseren^C 01:57, 25 April 2026 (UTC)

As laid out in the WT:GAN thread and supported by multiple editors, articles containing LLM-generated text merit an {{ai-generated}} tag which qualifies them for quick fail under criteria 3 of WP:QF. That, on its own, is enough to delist. If an editor wants to retain the rating, the onus should be on them to establish that the article has GA-level source to text integrity despite it containing - or even largely consisting of - PAG-violating text that qualifies for a quick fail. I am certainly open to such an argument but I do not see one presented here. NicheSports (talk) 03:35, 25 April 2026 (UTC)

Applying a nonsense tag without any actual noted issues with an article is not an excuse, actually. Otherwise anyone could slap a neutrality tag or a globalize tag on a GA and quickfail it because they felt like it. Again, unless you have actual, specific issues with the article, you're just making up nonsense otherwise. Silverseren^C 05:11, 25 April 2026 (UTC)

Regarding Linnaeus, he only lists a number of sources with quotes how those sources describe the species. The article, however, says that he provided a single, congruent description by himself, and then, in addition, cited sources "that describe the lichen similarly". That's a sonesensical error and something a human editor would just not do (a human can misinterpreted things, but a human would not provide two interpretations of the same list as we see here). --Jens Lallensack (talk) 04:13, 25 April 2026 (UTC)

It is 100% how a human would use the Species Plantarum as a source, because that is how Linnaeus' works are used. A human editor with actual experience in plant biology would 100% use it that way, because they understand what such a source means.

Tell me this, Jens. If a AI was used to entirely generate this info from this non-English source, what text did it use? Because, if you click the "Show Text" button on the right, you'll notice that in large part, the resulting OCR text is gibberish. It barely gets a couple Latin words in a row right before some error happens. Do you seriously think an AI will read proper words from, for example, "Scor7onera graca faxati)is& maritima"? No, the Latin was translated by human eye and utilized that way. Silverseren^C 05:11, 25 April 2026 (UTC)

I know Esculenta as a highly skilled expert editor, and worked with them on multiple articles. I refrain from believing that they would produce such a nonsense-paragraph on their own. Careless AI use is the only explanation that makes sense to me here. --Jens Lallensack (talk) 05:56, 27 April 2026 (UTC)