I can only tell you that I occasionally run my own texts through these detectors: with a 50:50 probability I am an AI ;-)) Seriously, I wrote about how it might be more useful to rely on interaction with authors and a sense of language in terms of form and expression. So far I've found a few accusations where I actually think the authors of the posts are beyond reproach. And that is unattractive. Add to that the fact that more and more (young) people will read and consume massive amounts of AI-generated texts. Their writing style will automatically follow suit at some point. At that point, at the latest, it will no longer be possible to distinguish between them.