The face is the first place where we look for a person. From a glance we infer identity, intention, emotion and even an instinctive measure of trust. That is why AI-generated faces are more than a technical advance: they intervene in one of the fastest and deepest mechanisms of human perception.
A study by the University of Milano-Bicocca, published in Scientific Reports on June 30, 2026, examined this relationship through electroencephalography. Thirty participants looked at 440 images, half real and half produced by generative systems. They were asked to judge the origin, familiarity and attractiveness of the faces while 128 sensors recorded brain activity.
The results reveal a split between what we consciously decide and what the brain processes. Participants correctly identified only 33 percent of synthetic faces as artificial—less than would be expected by random guessing. Yet neural signals still distinguished generated images from photographs, activating networks linked to recognition, memory and emotional evaluation in different ways.
Perhaps the most striking result concerns trust. AI faces were judged more familiar, attractive and trustworthy. Not because they resembled people the participants actually knew, but because generators tend to produce regular, prototypical physiognomies: statistically average, symmetrical faces with fewer individual irregularities. The machine does not simply create a believable face; it can create the kind of face our perceptual system finds easiest to accept.
For cinema, advertising, avatars and digital characters, this quality is extremely powerful. A synthetic face can win a viewer’s trust immediately without possessing a real history, body or identity. An aesthetic choice therefore becomes a psychological choice as well: how reassuring should a character be, and when does that familiarity risk becoming manipulation?
The study does not prove that the brain contains an infallible deepfake detector. The experiment used static images and a limited group of people, and a neural difference does not automatically translate into a conscious decision. It does show, however, that photographic appearance can no longer serve as proof of authenticity and that our sense of trust may be influenced before rational evaluation begins.
The new frontier of the artificial face is therefore not visual perfection. It is the ability to seem immediately close, familiar and credible. Once synthetic imagery crosses the uncanny valley, it does not stop producing effects; it moves them from obvious discomfort toward a quieter form of persuasion.