ANU Study Reveals New Method to Spot AI-Generated Faces

Researchers at the Australian National University (ANU) Emotions and Faces Lab, under the guidance of Associate Professor Amy Dawel, have introduced a training technique aimed at enhancing the identification of AI-generated faces. Rather than concentrating on obvious flaws such as extra fingers or levitating earrings, the findings published in Scientific American on June 30, 2026, emphasize six perceptual traits: distinctiveness, memorability, proportionality, symmetry, attractiveness, and expressiveness. Faces produced by models like StyleGAN3 are generally more symmetrical, proportional, and appealing, yet they lack expressiveness and memorability. ANU Honors student Tanya Georg conducted training sessions that enabled some participants to reach 100% accuracy. The study underscores the need for educational tools as AI-generated imagery evolves rapidly.