The Ultimate Undetectable AI? A Bot That Blunders!

Hey everyone, I had an interesting thought about AI-assisted play and how a bot could, in theory, avoid detection.

Instead of playing like a perfect 3200+ engine every game, what if a bot intentionally mixed in occasional human-like mistakes? Imagine it playing at a 94%+ accuracy level most of the time, but strategically throwing in blunders, inaccuracies, and even losing some games—just like a strong yet imperfect player.

This kind of AI would be way harder to detect than a bot that just plays like Stockfish all the time. It would win most games but still have believable losses, making it nearly indistinguishable from a talented, streaky player.

Not suggesting anyone actually do this, of course—just a fun thought experiment. What do you guys think? Could an AI like this fly under the radar? How would Chess.com even catch it?