If someone is using AI effectively, there's often no way to tell that they're using AI at all. Toupée fallacy etc.