Loading the Elevenlabs Text to Speech AudioNative Player...

TL;DR: AI-generated food photos went viral last week for all the wrong reasons. Shrimp that look like pool floats, burgers with cheese that behaves like lava, a menu item at a café that turned out not to exist. If you run a restaurant, this is the week to check what’s actually on your menu boards, your website, and your delivery-app listings.

I’ve spent the last two years telling restaurant owners that AI is worth their time. I still believe that. But “AI slop” made it onto real menus this month, people noticed, and the reaction was not kind. So let’s talk about what happened and where the line actually is.

What happened

Around the start of September, a wave of posts started circulating showing menus and storefront signs with obviously AI-generated food. TechCrunch covered the pattern on September 3, and the examples were rough: burgers with “cheese so bubbly and melty” they looked like modern art, shrimp that appeared “to have been genetically modified to eat their own tails,” bagel sandwiches too symmetrical to be real.

The part that should worry you most came from a customer who tried to order something off a café’s menu and found out the dish didn’t exist. The image had been generated in ChatGPT, and nobody had checked it against the actual kitchen.

The businesses involved were mostly small, family-run places. Not chains. Owners trying to get a menu out the door on a tight budget, using the tool that was in front of them. I understand exactly how that happens.

Why it looks wrong (and why customers can tell)

There’s a real explanation for why AI food looks the way it does. Image models learn what “a burger photo” is from millions of existing burger photos, and a huge share of those are from big chains. So the model reproduces that glossy, over-lit, perfectly round look. Then, every time you ask it to tweak the image, it drifts further toward smooth and symmetrical. Reality Defender’s CTO described it as “almost like an alien trying to make a pizza without understanding its core principles.”

Researchers at the University of Duisburg-Essen in Germany found something worse: AI food images that are almost realistic trigger more disgust than images that are obviously fake. It’s an uncanny valley effect, and it’s amplified by the fact that a lot of people are already suspicious of anything that smells like AI.

Translation for a restaurant owner: a slightly-off AI photo doesn’t read as “nice photo.” It reads as “something is wrong with this place.”

The delivery apps are pushing this too

This isn’t only owners freelancing in ChatGPT. DoorDash offers “Background Enhanced Menu Photos” that use AI to clean up photos you upload, and Uber Eats says it uses AI to “detect and enhance low-quality food images” including lighting, resolution, framing, and plating. Both say they don’t change what the food looks like. In my experience the enhancement tools are fine when they start from a real photo of your real dish. The trouble starts when there’s no real photo underneath.

Where I’d draw the line

Here’s the rule I give clients, and it’s held up through this whole mess:

AI can improve a real photo of your food. AI should never invent a photo of your food.

Concretely:

Try this today

Pull up your menu, your website, and your DoorDash and Uber Eats listings. Every photo should be a dish your kitchen makes, shot in your building. If you find one that isn’t, replace it with a phone photo in good natural light. An honest photo of real food beats a perfect photo of fake food every single time, and this month proved it.

Sources: TechCrunch, The Cool Down, TODAY.


Justin Peters runs Zoomies Media, an AI-powered marketing agency for local and hospitality businesses in Orlando. This post was drafted with AI assistance and edited by Justin. Read about how this site works.