Avatars versus the people: Photo-realism, spontaneous detection and trait inferences of digitised faces

In plain English

Scanned avatar faces were compared with passport-style photographs of the same people. Shown side by side, viewers told them apart 94.1% of the time, but shown one at a time accuracy fell to 87.4%, and only 28% of participants noticed on their own that they were looking at avatars at all. Judgments of attractiveness, dominance and trustworthiness carried over strongly from photo to avatar, which means a scan largely inherits whatever first impression the real face creates — useful if you are cloning a trusted presenter, and a problem if you assumed the avatar reads as a neutral stand-in.