The images that user Maggie Moda spotted outside city cafés share a recognizable weirdness. Croissant sandwiches flaunt comically perfect layers. A chopped-cheese hero sports cheese melted a little too uniformly over every piece of meat. One affogato sprouts random holes, and a burrito pockmarked with so many tiny perforations earned the internet nickname the “trypophobia burrito.” The pattern has become a running joke among critics who’ve branded the genre “food slop.”
Why Restaurants Reach for Generative AI
For small operators, the appeal is bluntly practical. Professional food photography costs money and scheduling time that many new restaurants simply don’t have. Generative AI tools can produce photorealistic images of dishes in minutes for a fraction of the cost — sometimes for free.
When AJ and Lyndsey Lozano opened Grind & Unwind on San Francisco’s Haight Street in May, they taped a temporary menu of AI-generated food images in the window while they got the business running and cared for three children under 5, according to SFGATE. The internet’s response was swift: a Reddit post titled “Yum, slop” drew more than 670 upvotes, with commenters comparing the focaccia to a “loofah” and a “lung.”
A Chicago deli offered essentially the same explanation. Ali Malik, manager of Chi-Town Deli, told Business Insider the AI-generated sandwich images were a stopgap while the restaurant waited for a new sign featuring actual photographs.
The Legal Precedent That Predates AI by Decades
Fake food in advertising isn’t new. In the 1965 Supreme Court case FTC v. Colgate-Palmolive Co., justices weighed a familiar trick: mashed potatoes standing in for ice cream because real ice cream melts under hot studio lights. The court recognized that advertising images didn’t need to be literal to be lawful.
The legal standard, as Harvard Law School professor Rebecca Tushnet explains, hinges on what a reasonable consumer understands an image to be claiming. “If a reasonable consumer would understand that the image was making actual claims about the composition or quality of the food, then that could be deceptive,” Tushnet told Fortune. But if the image simply conveys the general kind of product a customer will receive, “then it is not likely to be deceptive.”
Generative AI strains that distinction. Mashed potatoes masquerading as ice cream at least represent a real food photographed under real conditions. An AI image of a croissant sandwich with engineered-perfect layers represents nothing that ever sat in front of a camera. The image isn’t a stylized version of a real dish — it’s a synthetic rendering of a dish that doesn’t exist. For restaurants, that turns food photography from an art into a legal exposure.
What Happens Next
The backlash suggests customers may police the line themselves, at least for now. A viral post cataloguing awkward AI food ads can do more reputational damage in a weekend than a regulator can do in a year. Restaurants that leaned on AI images as a temporary measure — Grind & Unwind, Chi-Town Deli — have already said they’re moving toward real photographs.
But the economics of generative AI won’t disappear with the backlash. As the tools improve, the visual tells — repeating holes, unnaturally uniform ingredients — will likely fade, making AI food ads harder to spot and harder to mock. That could shift the burden from online shame to regulators and courts. Advertising law hasn’t yet caught up with synthetic food imagery, and the gap between what a reasonable consumer assumes an image represents and what it actually shows is only widening. For now, the most visible check on deceptive advertising isn’t the FTC — it’s a viral X post and a Reddit comment comparing bread to a lung.
— Isabella Morales, food desk, AXO News