From TechCrunch · Amanda Silberling · · 1 min
Why AI-Generated Restaurant Menus Look So Unappetizing
Generative AI menus are creeping into cafes, replacing real food with eerie, homogenized visual horrors.
In brief
Generative AI menus are creeping into cafes, replacing real food with eerie, homogenized visual horrors. Repeatedly editing AI-generated food art strips visual authenticity and repels customers through an uncanny valley effect. Originally reported by TechCrunch.
Eerily flawless menu art
Cafe menus increasingly feature illustrations that look unnaturally symmetrical, smooth, and visually off-putting.
“Generative AI menus have hit the restaurant business courtesy of models trained on a narrow, 'pleasing' aesthetic that feels wrong...”
Lovecraftian food horrors
AI generators frequently output strange monstrosities, like burritos with avant-garde melt patterns or shrimp eating their own tails.
“...every ice cream scoop is perfectly round, and where shrimp seem to have been genetically modified to eat their own tails...”
Trained on old data
Image models rely on massive training datasets that heavily mirror legacy fast-casual menus from years past.
“A lot of this stuff looks like a Chili’s menu from 2015, and there’s a reason for that. That was the corpus of work from which [the models] drew their function.”
The threat of model collapse
As AI content floods the web, models risk feeding on their own synthetic outputs, leading to severely corrupted generations over time.
“...when you feed the outputs from one model back into itself, eventually the inbreeding becomes too much, and the whole thing collapses...”
Convergence homogenizes visuals
Even without full model collapse, AI outputs suffer from convergence, blending distinct styles into a generic middle ground.
“Convergence is a bit less extreme, degrading the quality of an AI’s outputs without making it entirely useless.”
Shaving off the edges
AI image generators prioritize broad aesthetic pleasingness over real detail, flattening distinct culinary personalities.
“What AI is known to do both in images and language is to shave off the edges.”
Iteration ruins the image
Making tiny sequential edits to an AI menu item degrades the picture over time, making food look progressively less real.
“It seems that with each edit, the food images become a tiny bit more round and smooth.”
Consumers spot the fake
Customers instinctively notice when visual representations of food aren't real, often reacting with immediate skepticism.
“There’s just a sensibility that people sometimes find hard to articulate, but they kind of know it when they see it...”
The uncanny valley of food
Research indicates near-realistic AI food triggers visual disgust, eliciting stronger unease than obviously fake art.
“...images of food that looked almost real elicited more disgust and unease than images that were obviously fake.”
Seeing is no longer believing
Unappetizing menus signal a broader cultural shift where visual content can no longer serve as definitive proof.
“Seeing and hearing has always been believing... That’s no longer the case. The world has fundamentally shifted, for good or for ill.”
The headline
Repeatedly editing AI-generated food art strips visual authenticity and repels customers through an uncanny valley effect.
