Why AI-Generated Restaurant Menus Look Disturbingly Artificial

Something feels profoundly off about AI-generated restaurant menus, even if you can't immediately say what. The bagel sandwiches are too symmetrical, the ice cream scoops impossibly perfect, the burritos wrapped in cheese that bubbles like avant garde sculpture rather than lunch. Your instinct is correct—these images trigger a visceral discomfort that scientists now understand as a measurable phenomenon.

The Convergence Problem#

Generative AI models that produce menus are trained on vast datasets from which they extract patterns. When asked to design a fast-food menu, they typically reference widely available examples from chains like McDonald's or Wendy's, whose existing materials already share a uniform visual style. This creates a feedback loop: as AI-generated menus enter training datasets and get refined repeatedly, the models reinforce the same narrow aesthetic over and over.

Alex Lisle, CTO of Reality Defender, explained the mechanics this way: "A lot of this stuff looks like a Chili's menu from 2015, and there's a reason for that. That was the corpus of work from which [the models] drew their function." The phenomenon is called convergence—a milder version of "model collapse," where repeated outputs gradually degrade quality without rendering the system entirely useless.

Why Everything Looks Aggressively Pleasant#

AI training datasets are typically optimized for "pleasingness" and to avoid offensiveness. This filters out irregularities, asymmetries, and the imperfect details that make real food look appetizing. Restaurant menus have long featured images more polished than reality—a Big Mac styled by prop designers to maximize appeal. AI amplifies this effect dramatically.

Lee Rainie, Director of the Imagining the Digital Future Center at Elon University, described this as a flattening of character: "What AI is known to do both in images and language is to shave off the edges." The more a restaurant edits an AI-generated menu—tweaking prices, adjusting item names—the smoother and rounder the food becomes with each iteration, as demonstrated by experiments on social media.

The Uncanny Valley of Food#

Research from the University of Duisburg-Essen found that AI-generated food images trigger an uncanny valley effect—images that look almost but not quite real provoke more disgust than obviously synthetic ones. The more an image approaches photorealism without fully achieving it, the stronger the negative response.