AI and human faces evoke systematically different perceptual impressions.

Proceedings of the National Academy of Sciences (PNAS), a peer reviewed journal of the National Academy of Sciences (NAS) - an authoritative source of high-impact, original research that broadly spans the biological, physical, and social sciences. Humans trained to use global impressions of faces nearly double their accuracy at spotting AI-generated deepfakes. This approach beats training on local visual artifacts, which has shown little benefit. Gains persist in follow-up tests and scale in an online replication.