The sameness problem behind those unappetizing AI-generated menus

TechCrunch explores why AI-generated restaurant menus look uncannily similar, attributing the effect to training-data homogenization and model convergence rather than model collapse.

Restaurants are increasingly using AI-generated menus that exhibit a distinctive uncanny sameness, which Reality Defender's CTO attributes to models trained on a narrow, homogenized "pleasing" aesthetic. Training models on their own outputs risks model collapse, though the observed effect is convergence, which degrades output quality without total failure. Researchers at the University of Duisburg-Essen found AI-generated food images trigger an uncanny valley effect, eliciting more disgust than obviously fake images. Repeated edits to AI-generated menus progressively distort food imagery, fueling backlash against AI menus in the restaurant business.