Next time you're waiting in line for a sandwich, look up at the menu board.

If the sandwich looks a little too shiny, a little too perfectly layered, or faintly like it was designed by someone who has read about food but never eaten any, you may be looking at an AI-generated image.

What Diners Are Spotting

CBC News reports that diners in the Toronto area have been posting photos of menu boards with food they say "looks rubbery, surreal and seems to be created with artificial intelligence tools."

The examples are memorable. Shrimp shaped like doughnuts. Breaded shrimp and gyros "with suspiciously geometric textures." Posters at the Canadian National Exhibition this summer that visitors "assumed to be AI-generated." One Reddit user captioned a menu:

❝

"This looks like what aliens would put in a presentation about what Earth food is."

It isn't only Canada. NBC News described a shrimp scampi "whose shrimp look more like miniature life preservers," a breakfast burrito "with filling that looks more like something out of a horror movie," and "protein" options "that look like they're covered in bugs."

In San Francisco, a new café called Grind & Unwind put AI images on its menu. Redditors said the bread "looks like reptile skin" and compared the croissant to "alien larva," as Futurism reported from SFGate. Someone spray-painted "seriously" on the awning. The owners took the images down.

And the one that turns a joke into a problem: according to NBC, one X user said she tried to order an item advertised on a café's outdoor menu, only to be told it didn't exist. It had been made with ChatGPT.

A quick caution: most of these are suspected AI images, identified by diners rather than confirmed by the businesses. The pattern is real, though, and the reactions are too.

Why Restaurants Do It

Mostly, money.

Restaurants run on thin margins. In Restaurants Canada's most recent survey of some of its members, 41 per cent said they were operating at a loss or just breaking even. A professional food shoot costs money and time: book the photographer, wait for the date, wait for the images. An AI tool costs almost nothing and takes minutes.

Not every use is slop. CBC spoke to Fez Mian, a marketing manager at The Fancy Induced Burger, a Canadian franchise. He photographs the actual burger as it comes out of the kitchen, then uses ChatGPT to "enhance the texture and freshness." A person checks that no extra ingredients were added. "Within … five minutes, 10 minutes it's out the door."

That's editing a real photo of real food. It's much closer to what food photography has always involved than to a sandwich invented from scratch.

Is It False Advertising?

The marketing experts CBC interviewed are calmer about this than the comment sections.

David Soberman of Rotman says the problem is an image that's very different from the product, like a steak that looks an inch thick on the menu and arrives half that. But he says the same applies to traditional staging and editing. Both he and David Pullara of Schulich say AI imagery doesn't generally make misleading customers more likely, since most people already expect menu photos to look better than the real thing.

Pullara put the real issue neatly:

❝

"I don't think it's the AI that makes the image unappealing. I think the problem is asking AI for a photo and just not using a common-sense filter about [whether] this actually looks good."

The food stylists have a point too. Toronto stylist Zoe Rousseau says modern food styling mostly doesn't use fake food: it's careful ingredient selection, placement and lighting. And if an AI image "doesn't look like the food," it won't bring anyone in. Her advice to small restaurants: "better off taking pictures with their iPhone."

The Easiest AI-Literacy Game There Is

Most AI-literacy advice for kids is abstract. This one is right there on the wall, and it's fun.

Play "Could This Food Exist?" Next time you're out, look at the menu pictures together. Is this a real photo, a real photo that's been edited, or something that was never cooked at all?

Look for the tells. These are our suggestions, not the article's, but they're what the viral examples have in common:

  • Shapes that are too perfect. Shrimp as exact rings. Layers that line up like a diagram.

  • Patterns that repeat. Bread texture, sesame seeds or grill marks in a regular grid.

  • Textures that don't make sense. Meat that looks like leather, coral, or plastic.

  • Everything shiny. A gloss on food that nothing edible would have.

  • The plate test. When the food arrives, compare. How close was it?

Teach the three categories. This is the idea worth carrying into everything else a child sees online:

  1. A real photo of real food.

  2. An edited photo of real food, which is normal and has always happened.

  3. A generated image of food that was never cooked, which may or may not match what you get.

The difference between 2 and 3 is the difference between marketing and invention. It's the same skill a teenager needs for photos of people, places and news events, and practising it on a doughnut-shaped shrimp is a lot more fun.

Be kind about it. Many of these businesses are small, family-run and short on money. The lesson is about noticing, not mocking.

The Part Worth Remembering

A menu board used to be a picture of dinner. Now it might be a picture of what an AI imagines dinner looks like.

Your kids will spend their lives surrounded by images that were never taken with a camera. A sandwich board is a good, low-stakes place to start noticing.

Quotations verbatim. Most "AI slop" examples are identified by diners, not confirmed by the businesses. The café item that didn't exist is an X user's account as reported by NBC. The list of visual tells and the family guidance are ours.

Disclosure: drafted with Claude, made by Anthropic.

Want practical AI guidance for parents and educators every week? Subscribe: https://www.aibyage.com/?modal=signup&utm_source=beehiiv&utm_medium=newsletter