In one peer-reviewed study of 14 detectors, none achieved 80% accuracy, and one in five allowed AI-generated text to slip through undetected. The collateral damage is piling up: blacklisted journalists, students expelled from universities, and academics charged with misconduct.

