AI Models Overrate Facial Attractiveness, Preliminary Study Finds

A preliminary study has found that ChatGPT, Claude, Grok and Gemini can closely mirror how humans rank facial attractiveness, while assigning faces higher scores.

The findings point to a distinction between ranking faces and rating them. The models showed strong alignment with human judgments about relative attractiveness, but were more generous in their assessments.

That means agreement on which faces are more attractive does not necessarily translate into agreement on how attractive those faces are.

The study remains preliminary. The available summary did not provide details about its methods, the faces assessed or the size of the gap between human and AI ratings.