Surpassing the uncanny valley

The present study aimed to determine whether the human brain preserves sensitivity to the artificial origin of hyper-realistic AI-generated faces even when explicit, behavioral discrimination from real faces fails. By integrating behavioral validation with high-density EEG recordings, we sought to characterize the temporal dynamics and neural substrates underlying the processing of real versus GAN-generated faces across perceptual, evaluative, and familiarity-related stages.

Specifically, the study investigated whether implicit neural markers reveal systematic differences in the processing of artificial faces that are behaviorally indistinguishable from real ones, thereby dissociating overt recognition performance from covert neural evaluation.