I have been following the thread of nutildah, and at that thread he has shown some of those detectors which we users can use to detect AI generated text. The accuracy of those detectors is still not 100% confirmed because they can detect original text that has been grammar fixed with the help of those AI tools, however they are efficient at detecting many texts which can often be confirmed as AI generated texts.
Those tools are getting tough each day and I'm quite sure that the day isn't far when they will detect human generated text as AI generated text, and on that day we will have to find a better solution to detect such texts. However, as of now we can at least trust those detectors even if their accuracy is still unconfirmed.
Sometimes those texts may contain human response, but they have been polished with AI tools to fix grammar errors. I think the moderators know about this fact and that's why they don't delete all those posts because they have a human touch. However, it's always better to work on our grammar rather than relying on those tools because they can make us suffer in long term.