The first warning signs were visual: pasta scattered with inexplicable black dots, shrimp that looked wrong on close inspection and pastries topped with purple rather than white sugar. What has become known as AI food slop is increasingly appearing on restaurant signs and menus, replacing the glossy food photography designed to make customers hungry. One account of the trend called the resulting images “far less appealing than whatever food porn was.”¹

The problem has spread from London to New York as fast-food businesses turn to generative AI for marketing that can be produced in minutes. The giveaway is often an uncanny excess of detail: gray burritos, malformed sandwiches and repeated holes or specks where sesame seeds or texture should be. AI expert Jon Krohn says many of the viral failures likely come from older, free models used with simple prompts — not from the best current systems. “Humans are unusually sensitive to images of food that don’t look real,” he said, because judging food quality matters at an evolutionary level.²

For restaurant owners, however, the calculation is blunt. Professional food photography costs money; AI is cheap, immediate and endlessly editable. Alexandra Coronel, a manager at Nature Valley Market in New York, used ChatGPT images to promote paninis and bowls but worries they could alienate customers. At Vinny’s Gluten Free Kitchen, worker Enzo Bar framed the choice as competitive necessity: “People want the real thing without paying for it.”

Diners are responding by trusting what they can see in front of them, rather than the menu art. Customer Devan Currao said she would prefer an unpolished real photograph to having to guess whether an AI image resembles the dish. Food photographers share that concern, calling the practice potentially misleading when the picture has no direct connection to the product.

The distortion may not last: improving models could make synthetic food harder to detect. But the dispute is not only about pixels. Even photographer Lorenzo Franco, who uses AI to enhance his own work, sees a limit to the machine’s role: an AI agent is “never going to be hungry.”