I think every moderator on every platform is struggling with this issue, and no one has succeeded so far, so it doesn’t seem that easy.

I think a simple solution (and one that eventually every content platform will have to adopt) is to allow users to tag AI-generated spam. I think that a few years from now this feature should be the norm, like existing basic features on forums such as upvote, downvote, favorites, hide, etc. I know this will require much more development effort than simply blocking new accounts from posting at all. But on the other hand, you can’t block new accounts forever.