Food delivery apps Deliveroo, Uber Eats, and Just Eat are being flooded with fake AI-generated photos of dishes on restaurant menus across Britain.
Unnaturally perfect burgers, strangely spherical rice, and physics-defying snacks are appearing on menus nationwide. Reporters from the Daily Mail ordered food from several of the worst-offending takeaway outlets to test how the artificial photos compare with the actual meals delivered.
While a disappointing meal may be minor, experts and food photographers warned that the practice amounts to dishonest false advertising that threatens to destroy public trust in the restaurant industry.
Some of the artificial menu photos clogging up popular ordering platforms are strikingly obvious fakes upon examination. Image checker tools confirmed the artificial origins of photos that featured glaring visual glitches.
At Pasta Bowl and Feast, a takeaway restaurant listed on delivery apps, an image of mozzarella sticks displayed cheese pulled into a strange loop resembling an M.C. Escher painting. Modern image generators usually avoid such classic flaws, making its appearance on a commercial menu surprising.
When the mozzarella sticks arrived, they were recognizable and edible, but looked blander, less crisp, and significantly less appealing than the menu photo had suggested.
Reporters test takeaway orders
A similar pattern emerged at Mimi Noodle and Tap, a nearby Chinese takeaway on Uber Eats that had zero customer reviews. The restaurant menu displayed glistening sauces, steaming hot dishes, and vegetables chopped to impossible uniformity.
The ordered Chinese Pork Ragù Rice Bowl was reasonably close to its photo and proved to be a good lunch, although it lacked the high polish shown in the digital menu image.
Other orders produced far more severe disappointment. Five Rivers Rotisserie advertised a rotisserie chicken burger with a menu photo that already looked like an artificial mess. The delivered burger arrived flat and floppy, wrapped in a soaked piece of bread that felt as though it had been dropped in a mayonnaise-flavoured puddle.
How to spot artificial menu images
Consumers seeking to identify artificial photos on ordering apps can watch for several warning signs. Downloading an image and running it through an online detection tool such as SightEngine can confirm whether a photo was generated by artificial intelligence.
- Look closely at small details and finer points where software frequently makes mistakes.
- Question whether the food looks unnaturally perfect, as artificial tools exaggerate natural textures and symmetry.
- Check the lighting and shadow alignment, which generative tools struggle to keep consistent.
- Examine backgrounds for odd stretching, warping, or unusual blurring.
Why fake dishes look better
Research indicates that consumers frequently prefer artificial food photos over genuine photography. Behavioural scientist Dr Giovanbattista Califano, from the University of Naples Federico II, studied the visual impact of computer-generated food imagery.
"AI-generated food images are consistently rated as more visually appealing than real photographs," Dr Califano said. "The reason is that generative models tend to idealise food: images come out glossier, more symmetrical, better lit and more flawless than a real dish photographed under normal conditions. So an AI image can look 'better than real' simply because it strips out the imperfections a camera would normally capture."
Scientists classify artificial food images as superstimuli, meaning they carry exaggerated traits that heighten human responses. Psychologist and cognitive neuroscientist Dr Alexander Diel, from LVR University Hospital Essen, noted that food advertising has long used photo editing to create superstimuli, but generative tools allow anyone to create them effortlessly.
Dr Diel stressed that generating a dish from scratch differs fundamentally from editing a photo of real food. He explained that an AI-generated image of a food served by a specific restaurant will likely not correctly represent the actual food dish, as the AI is not trained on that restaurant's food images, meaning customers will not receive what was pictured.
Ordering from a restaurant named SPARTA!!! on Just Eat produced similar results. The takeaway offered Falafel Bombs, described as golden on the outside and soft inside, accompanied by a glossy picture. The delivered meal consisted of a cardboard cup containing brown, cardboard-like pucks.
At Taste of China, which raised red flags by operating out of a pizza restaurant, the menu featured wax-like dishes and mysterious spheres. Most incongruously, the menu pictured a stack of fluffy American-style pancakes listed as an accompaniment to roast duck.
An order of cheap pancakes and Sweet and Sour Prawn Balls, which looked like Ikea meatballs stuffed with cake batter, took an hour of phone calls to track down. The delivered prawns spilled liquid oil across the desk when opened, while the duck pancakes arrived as a small, sad plastic bag of plain wraps.
Industry fears over lost trust
Commercial food and drink photographer Brent Herrig, based in New York, said restaurants are turning to artificial images due to cost and appetite. He explained that social platforms demand new images constantly, while restaurants running on thin margins cannot afford weekly photo shoots, allowing artificial imagery to win over rushed phone photos.
Herrig said using artificial intelligence to enhance lighting or expand backgrounds is acceptable, but emphasized that a clear line is crossed when an image depicts a specific dish that was never cooked. He warned that food photography relies on an implied promise that someone stood in a kitchen and prepared the dish, warning that removing that truth at scale discounts the entire category.
Los Angeles-based food photographer and director Lindsay Kreighbaum called using an artificial image totally misleading if it is not a photo. She expressed concern over false advertising, stating that customers should not have to use their imaginations when deciding on a meal and warning that dishonest images will sever trust between customers and businesses.
Rules and platform responses
Major delivery platforms maintain policies regarding menu imagery accuracy. Uber Eats mandates that photos accurately represent a single menu item and prohibits customer-uploaded artificial pictures, though it did not provide a statement in response to inquiries.
Deliveroo states that pictures may be rejected for obvious signs of image editing and employs automated filters to detect fakes. Deliveroo did not issue a statement regarding the findings.
Just Eat requires restaurant partners to use real photos owned by the business and forbids artificial images, also using automated filtering tools. A spokesperson for Just Eat said the company supports tens of thousands of independent restaurant partners to ensure menu accuracy, including funding local professional photographers, adding that non-compliant images highlighted in the investigation have been removed.
In the United Kingdom, the Food Standards Agency requires food businesses to market and advertise food accurately without misleading consumers regarding quality, ingredients, or production processes, though current regulations do not explicitly ban artificial intelligence. All referenced takeaway outlets were approached for comment.
