That is true! Good point. They like us to think they are trying to deal with AI and cannot reliably identify it, but if they found a way - or at least tried harder - they could eliminate one source of leakage, leaving more money for real writers.

MSN notably stops people commenting on news stories if they use trigger words, and there are a lot of trigger words. Spam messages could easily be flagged as trigger words, and so could anything that triggers an AI detector at more than 50% AI generated.