I don't necessarily agree that all of the accounts in that list are consistently posting human generated content or that there's no use of AI going at all - even among longstanding community members. But I do agree that a single report from a single site is not enough to be anywhere near certain.

I have, informally, adopted rules something like this when I curate in Popular STEM:

1. Give people the benefit of the doubt. If I'm not confident that it's AI generated, I assume it was created by a human.
2. Get feedback from multiple detection sites (currently, I use 4 of them)
3. Look for patterns of repeat behavior. If multiple AI checkers are triggering, consistently, on multiple posts by the same author, it's much more likely to be correct.

Also, I agree with this comment from @chriddi. The use of online translators makes it even more difficult to be sure, because they probably use the same sort of statistics that AI large language models use in order to perform the translation.

A final point that I have grappled with: if we can't tell, does it matter? The point of content creation is to bring human attention to the blockchain. It could be argued that rewards should go to the posts that do that best - perhaps without regard for whether they were created by humans or not. I'm not ready to make that argument. It just feels wrong. But I don't have a strong refutation, either.