AI menus’ sameness problem turns off diners

AI-generated menus are appearing in restaurants, but customers often find the food images unsettlingly flawless and artificial, triggering a visceral sense that something is wrong. Experts say AI models trained on a narrow, "pleasing" aesthetic produce homogenized visuals, like perfectly round ice cream scoops, that mimic popular chain menus from years past. Repeated edits can further degrade images, creating an "uncanny valley" effect that disgusts viewers. This aversion stems from model convergence, where AI outputs reinforce a bland sameness, and reflects broader concerns about AI's impact on trust in visual evidence, as detection startups and researchers study the phenomenon.