Restaurants, cafes and food brands are increasingly using AI-generated images to advertise their food, producing a wave of bizarre and unappetising results, from noodle-like pastries and stringy chicken to burgers and burritos riddled with strange lumps and holes. Experts say this "AI slop" phenomenon reveals fundamental weaknesses in how image-generating AI models work, which matters as more businesses turn to cheap AI visuals instead of real photography to promote products people expect to look genuinely appealing.

Specialists explain that most leading image generators use diffusion, building images from noise by establishing coarse structures first and adding fine detail afterwards; if the basic structure is wrong early on, later texture details compound the error, similar to how AI-generated people end up with extra fingers. Diffusion models are also particularly weak at rendering thin, repeating or terminating shapes such as noodles, strands and seeds, causing artefacts to bleed into places with no culinary logic, while the AI's lack of true understanding of what food actually is, gained only from statistical patterns in training data, leaves it unable to judge whether an image looks physically or appetisingly correct.