AI-generated faces — known as deepfakes — are flooding the internet, often used to amplify scams involving fake romantic partners or supposedly trustworthy investment contacts. New research from the University of Florida shows that these synthetic faces can actually appear more authentic than photographs of real people, a phenomenon the researchers call "deepfake hyperrealism."

According to a study published in October in the journal Social Cognitive and Affective Neuroscience, the effect is linked to a specific brain region called the anterior insula. This area, located deep inside the brain at roughly the level of the temples, acts as a central hub that helps the brain recognise important or potentially suspicious information. When participants viewed deepfake faces, their anterior insula showed reduced activity compared to when they viewed real photos, potentially explaining why the fake images seem so convincing.

"Fake faces are being perceived as more real than real faces," said Natalie Ebner, Ph.D., a professor of psychology at the University of Florida who led the study. "You're actually worse than just guessing, because you're actively being drawn in the wrong direction."

The research team assembled 80 images of faces — half real photographs and half generated by AI — and showed them to 126 participants ranging from 18 to 89 years old. Participants rated the reality of each face while undergoing brain scans that tracked activity in the anterior insula.

Both younger and older adult groups rated the AI-generated faces as more real than the real photographs. But older adults showed a significantly stronger bias, rating the deepfakes as even more realistic than their younger counterparts did. The researchers noted that activity in the anterior insula did not differ significantly between the age groups, so the reason for the stronger bias in older adults remains an open question.

"All adults have hyperrealism bias, not just older adults. But older adults have it more," Ebner said.

Ebner explained that AI-generated faces likely appear more real because they are generated as an average of many different faces, which triggers the brain's preference for familiarity. "We know that when you average across faces you end up with something that has less distinct features, is more attractive and represents more of the prototype," she said. "That makes these faces easier to process and more pleasing."

The findings have particular relevance for university students, who spend a significant amount of time online and may encounter deepfake profiles on social media, dating apps, or professional networking platforms. While younger adults showed a weaker bias than older adults, the study found that all age groups are susceptible to the effect.

Ebner said that AI-detection tools can help people assess the genuineness of photos online, but that the real advantage may simply be in raising awareness about the hyperrealism bias. This is especially important for older adults, who are often targeted for scams involving fake people using AI-generated faces, with many not even knowing that deepfakes exist.

"Identifying that this potential deception scheme is the first step to an intervention," Ebner said. "We can now work on awareness that if something looks too good to be true, think twice before you go along with it."