DEV Community

Auton AI News
Auton AI News

Posted on Originally published at autonainews.com

Uber Eats Catches Five Pizza Restaurants Using the Same AI Food Photo

Key Takeaways

  • Researchers found “almost real” AI-generated food images trigger more disgust and unease than obviously fake ones.
  • AI image generators produce visually homogenous food photos, resulting in unnaturally symmetrical dishes and blurred ingredients.
  • Diners prefer menus with no photos over AI-generated ones and some avoid restaurants using synthetic imagery. Five pizza restaurants on Uber Eats were recently caught using the exact same AI-generated food photo. That kind of slip is becoming harder to hide, and diners are starting to walk away because of it.

The Uncanny Valley on Your Menu

Researchers at the University of Duisburg-Essen in Germany found that AI-generated food images sitting in the “almost real” zone trigger more disgust than images that are clearly fake. The more convincing the illusion, the worse the reaction. Alex Lisle, CTO of Reality Defender attributes this to how image generators reproduce visual patterns without any underlying logic for how a dish should actually look, producing what he called “Lovecraftian food horrors”: shrimp that appear genetically modified, ice cream scoops with impossible symmetry, ingredients dissolving into each other.

Adaeze Onejeme encountered this firsthand at a Hollywood night market, where several vendors’ menus were filled with AI-generated images that left her questioning whether the food amounted to false advertising. Her reaction is not unusual. Online, the shorthand “AI slop” has spread quickly among diners who recognise the aesthetic even when they cannot fully explain what is wrong with it.

Why Everything Looks the Same

The sameness problem starts in the training data. Image generators like Midjourney and ChatGPT learn from datasets curated toward pleasing, broadly acceptable visuals. Over time, that optimization trims the unusual, the regional and the genuinely appetising in favour of a narrow ideal: every dish gleaming, every portion symmetrical, every colour saturated just enough. The result is a kind of visual averaging that makes one restaurant’s AI menu look almost identical to another’s.

A similar dynamic shows up in AI-generated text. Research on AI writing assistance found that when Indian participants used AI tools, their writing shifted measurably toward American stylistic norms, flattening cultural specificity in the process. The pull toward dominant training patterns affects images and words alike, and in food marketing, where a restaurant’s distinctiveness is part of what it is selling, that flattening is a real problem.

What Restaurants Are Doing Instead

Brendan Sweeney, CEO of restaurant marketing company Popmenu, has noted cases of identical AI images appearing across multiple pizza restaurants on Uber Eats and argues that using AI to enhance real dish photos is a sounder approach than generating images from scratch. Some platforms are already moving to remove AI-generated images from menu listings altogether.

The broader lesson for restaurants is practical: back-of-house AI applications, supply chain forecasting, recipe optimisation, inventory management, carry far less reputational risk than customer-facing creative work. When diners are the audience, synthetic perfection tends to backfire. As of August 2026, many online commenters say they would rather see no menu photo at all than an AI-generated one, and some say a fake-looking image is enough reason to take their business elsewhere.


Originally published at https://autonainews.com/uber-eats-catches-five-pizza-restaurants-using-the-same-ai-food-photo/

Top comments (0)