It was the kind of moment that makes you want to put down your phone and laugh. We have grown used to artificial intelligence writing emails, planning holidays, and even recommending films. But when you open a food delivery app because you are hungry, you expect something simple: a suggestion for something tasty, warm, and reliable. That is not what happened with Just Eat. The company, one of the biggest food delivery services in the world, found itself at the centre of an awkward public embarrassment. A new AI feature on its app was supposed to help customers choose meals. Instead, it began producing bizarre, unhelpful, and occasionally alarming food suggestions. Screenshots were shared all over social media, with people poking fun at the technology. Eventually, Just Eat was forced to say sorry. It admitted the artificial intelligence had not been working properly and said it was taking the feature offline. By that point, the damage had been done. The internet had already turned the failure into a viral joke, and another corporate AI experiment had become an unforgettable public relations disaster.
The exact details of the mistakes were both funny and worrying. According to the BBC report, the AI was not a human chef who had a bad day. It was a machine-learning system that had been trained to understand natural language and provide answers in a conversational way. In theory, it should have been able to answer simple questions like “What should I order tonight?” or “What goes well with pizza?” In practice, it did something far stranger. It made up responses that no cookbook would ever contain. The AI invented dishes that did not exist. It suggested combining ingredients that should never be combined. It offered advice that sounded confident but was completely wrong. These are what technologists call hallucinations: the AI generates fluent, believable text, but it has no understanding of whether its words are true. For a food app, this was a new kind of nightmare. People use Just Eat to satisfy a basic need, and the AI’s suggestions made no sense. It seemed not to understand food, cooking, or even the simple joy of a good meal.
Social media reacted the way social media usually does: with speed, humour, and no mercy. Users on X, Instagram, TikTok, and Reddit began sharing screenshots of the AI’s strangest moments. Some jokes were gentle. People compared the AI to a friend who had eaten one too many gummy bears. Others were sharper, suggesting that the company had clearly not tested its new system before letting it loose on millions of hungry customers. Fake menus began to appear online. Memes imagined the AI as a chaotic chef who had never tasted food. Restaurants and other brands joined in, trying to grab some of the viral attention. At first, it was all quite funny. Then, quietly, a more serious thought began to emerge. If an AI cannot reliably recommend a meal, what else is it getting wrong? People with allergies, dietary restrictions, and young children started to worry. The jokes continued, but the underlying anxiety grew. For a few hours, Just Eat was the most talked about food app on the internet, and not for a good reason. It was a painful lesson in how fragile brand reputation can be in the age of screenshots.
Just Eat’s response came quickly, even if it was slightly embarrassed. The company apologised sincerely, acknowledging that the AI had not lived up to its promises. It said the feature had been part of a small test, designed to help customers discover new dishes. It was never meant to replace human judgement or provide official nutritional advice. The company said it was switching off the tool immediately while it reviewed what had gone wrong. It also reminded customers that real people still handled the ordering process and that the AI did not affect the food being prepared or delivered. That was an important point, but it did not override the images already circulating. Social media has a long memory. A public apology can help, but it cannot delete the laughter or the skepticism. Still, the response was careful and professional. Just Eat did not try to blame the customers or hide behind technology. It accepted responsibility, promised to improve, and said human oversight would remain in place. In the corporate world, that is about as graceful a retreat as one can expect.
This story is bigger than one bad chatbot. It is the latest reminder that generative AI, for all its remarkable talents, is still fundamentally an imperfect technology. It does not reason like a human. It parrots patterns learned from enormous amounts of data. When it does not know the answer, it makes one up. This has caused problems across many industries. Airlines have had chatbots invent policies that never existed. Search engines have told users to put glue on pizza. Even some of the largest tech companies have watched their AI assistants produce offensive or nonsensical replies. Just Eat is simply the newest name on a growing list. But food is different. A chatbot that gets a movie fact wrong is annoying; a food assistant that gives dangerous advice is frightening. There is a real risk that somebody follows a bad suggestion and ends up with an upset stomach, an allergic reaction, or something far worse. That is why companies must be cautious. The technology should be tested, filtered, and supervised. It should never be released into the wild just because it is impressive in a demonstration.
In the end, the Just Eat story is a very human one. It is a story about hungry people, enthusiastic engineers, and a company that took a risk. It is also a story about how easily technology can embarrass us. The AI was not trying to be malicious, but it was trying to be helpful, and it failed. For customers, the lesson is simple: treat AI suggestions like advice from a stranger at a bus stop, not like gospel. If a recommendation sounds ridiculous, trust your instinct and order something normal. For companies, the lesson is harder. AI is powerful, but it needs guardrails, warnings, and human checks. Just Eat has apologised and promised to do better. We hope it keeps that promise. The next time we ask an app for a dinner idea, we want a helpful answer, not a viral joke. And if the choice is between a machine’s imagination and a trusted local restaurant, perhaps the wisest decision is to order the same thing you always order. Sometimes the old ways are still the best ways.