The Heated Rivalry fandom has such a problem with people posting AI generated/assisted fanfic and lying about it that some people have figured out an AO3 site skin that detects some usages of it and alerts you when you open a fic. It finds actual HTML markers in the code, so this is for sure LLM usage, not just tentative accusations based on style and whatnot. You can read about it in a twitter thread here, or the google doc!
It only works if the "author" used Claude and copied directly from Claude into AO3, so if they edited it elsewhere before posting, or used a different LLM it doesn't work, but still, I think it's very useful. Of course now that that's out in the open, people trying to hide their LLM usage will just delete the Claude markers, but you can at least catch the laziest ones!
Also fun fact the "author" of a popular work in the HR fandom left the fandom, crying harassment over "false" AI accusations, and others privated or deleted their works in support of them. But it turns out they had the Claude CSS tag in their works, proving they did use AI. So.
