It's the false positives that make it useless. Even if it's generally very good at detecting AI, the fact that it can and does throw false positives (and pretty frequently) means that nothing it says means anything.