AI detectors are failing to keep up with increasingly realistic deepfakes. Researchers have found that structured training can dramatically improve people's ability to identify AI-generated faces. The findings highlight the need for human vigilance alongside automated detection.