Restaurants' AI Menu Photos Are Sparking Backlash, Even Vandalism

Someone spray-painted graffiti over the storefront of Grind & Unwind, a San Francisco cafe, after it hung menu signage featuring AI-generated images of its food. Owner Lyndsey Lozano says she spent roughly $700 in July repainting the damage, an unusually physical response to what has otherwise been a mostly online controversy.

The backlash isn't limited to one cafe. In Denver, a manager at a Jamaican Jerk and Barbecue pop-up used ChatGPT to generate menu images rather than hire a graphic designer. When nurse Jill Sennett posted the results on X, more than 500 people reshared them, largely to mock how unappetizing the food looked.

The complaints aren't just aesthetic squeamishness. According to Reality Defender CTO Alex Lisle, generative models trained on massive image datasets tend to produce telltale artifacts: ice cream scoops that are too perfectly round, seafood with anatomically impossible proportions. These aren't rendering glitches so much as statistical averages baked into how the models learned to depict food.

TechCrunch reporting points to a deeper pattern behind the criticism: a "sameness problem." Because so much of the training data skews toward the visual style of major fast-food chains, AI-generated menu imagery across unrelated restaurants tends to converge on the same glossy, McDonald's-or-Burger-King aesthetic, regardless of the cuisine actually being served.

A model trained largely on chain-restaurant photography will keep producing chain-restaurant photography, no matter what dish it's asked to illustrate.

That mismatch is precisely what stung the Jamaican barbecue pop-up: images meant to represent a specific cultural cuisine ended up looking generic and off-putting instead, which is part of why the backlash online was so pointed.

None of this has stopped adoption. A 2026 National Restaurant Association report found 26% of restaurant operators now use AI in some capacity, spanning marketing, inventory, scheduling, menu optimization and order-taking, not just images.

That gap between operational adoption and visible failure is the real story. Restaurants are increasingly turning to AI tools to cut costs on tasks like design work, but the same convergence that makes these models cheap and fast is also what makes their output easy to spot, and easy to mock, when it's the first thing a customer sees. <tldrEn"> <targetAudiences">["BEGINNER","INTERMEDIATE"] <categories">["Ethics","AI Agents"] <suggestedSections">[] <suggestedTags">["ai-generated-images","restaurant-tech","generative-ai","consumer-backlash"]