Restaurants have been serving up fake food ads for decades—AI slop is just the latest special

New York has apparently become a test kitchen for AI food ads. A post by X user Maggie Moda, which has drawn 2.8 million views, showed sidewalk signs outside delis and cafés with croissant sandwiches, an affogato, and a chopped-cheese hero rendered in that uncanny, overclean style only generative AI seems to love. The reactions were harsh, but the business logic is easy to see: for small restaurants, AI can stand in for a pricey food shoot and get the menu in the window fast.

That convenience comes with a strange cultural hangover. Restaurants have been making food look better in ads for decades, but AI pushes the trick into new territory. It doesn’t just flatter the fries. It can invent a photorealistic meal that never existed, then use that image to sell a real sandwich to a real customer. That’s where the old question gets sharper: how far can the ad drift from the plate before it crosses into deception?

The internet has found its answer in vibes and ridicule. Some of these images are so tidy they start to look wrong at a glance: rows of identical holes, cheese melting with no clear source, croissants with absurdly neat layers. One burrito became known online as the “trypophobia burrito.” The joke name is rude, but the reaction is telling. People are less bothered by imperfection than by a fake perfection that feels like it’s trying too hard.

The actual restaurant cases behind the backlash were much less dramatic. Grind & Unwind in San Francisco put AI-generated food images in a temporary window menu while it opened in May, and owner AJ Lozano said the couple were juggling a new business and three children under 5. Chi-Town Deli in Chicago gave the same basic answer: the AI sandwich pictures were temporary until a new sign with real photos arrived. The internet saw slop; the owners saw a stopgap.

Legally, the line still looks familiar even if the tools don’t. A 1965 Supreme Court case, FTC v. Colgate-Palmolive Co., already wrestled with fake food imagery, including mashed potatoes posing as ice cream. Harvard’s Rebecca Tushnet says ads can use unreal images as long as they don’t mislead people about what they’re buying. Libby O’Neill of Loeb & Loeb goes a step further: an AI label doesn’t fix a false impression. The simplest rule is the bluntest one — sell what you show, and show what you sell.