AI-Generated Restaurant Menus Show Eerie Sameness, Triggering Widespread Unease
Generative AI models trained on narrow datasets produce food illustrations that appear uncannily perfect and homogeneous, causing restaurants and consumers to recognize and reject them.
What happened
Restaurants have begun using AI-generated images for menus, but the results exhibit a consistent aesthetic problem: food appears perfectly symmetrical, unnaturally smooth, and oddly uniform across different establishments. A user named Labtec demonstrated on X that iteratively editing an AI menu 100 times caused food images to become progressively rounder and smoother. Researchers at the University of Duisburg-Essen found that AI-generated food images trigger an "uncanny valley" effect, producing more disgust and unease than obviously fake images. Multiple experts—including Alex Lisle of Reality Defender and Lee Rainie of Elon University's Imagining the Digital Future Center—attribute this to AI models trained on large datasets that converge on similar visual patterns.
Context
The problem stems from how generative AI models identify patterns in training data. When models train on popular fast-food chains' menus or food advertisements—which already share similar professional styling—their outputs reinforce that same narrow aesthetic. This effect intensifies through "convergence," where AI-generated content re-enters training datasets, further homogenizing future outputs. The issue differs from catastrophic "model collapse" but still degrades output quality by eliminating visual variation. Additionally, AI models optimize for "pleasingness" and inoffensiveness, which compounds the smoothing and homogenization of images. This has broader implications: as AI-generated content becomes harder to distinguish from real imagery, it undermines the evidentiary standards that courts and verification systems have traditionally relied on.