A few thoughts:
First, it is almost comically mean to use the results of a project collecting AI tells to get LLMs to not sound like that. Like, of all the digital labor exploitations of AI, this might be the pettiest.
Second, AI detection is hard, and for all my concerns with AI, I think this is another good example of why policing its use can do more harm than good. I don’t blame the Wikipedia community for doing this project, but I would never recommend this approach in a classroom.