AI hallucinations — false or invented information produced by artificial intelligence systems — have begun to create serious problems in the customer service sector, with employees in restaurants, cafes, cinemas and shops facing customers who blindly trust ChatGPT or other generative AI instead of the professionals themselves. This phenomenon is dramatically reshaping the relationship between consumers and employees, while raising serious questions about the safety, reliability and limits of technology in everyday life.
See also: Verizon lays off 3,000 employees – AI in customer service
Madison, a waitress at a Michelin -starred New York restaurant , describes typical incidents that could be dangerous. Every time she greets new customers, she asks them about possible allergies. But recently, she has encountered dangerous situations: customers who claim to have a seafood allergy are ordering fish with shellfish broth because ChatGPT told them the dish is allergen-free. “ They tell you they’re allergic to shellfish, and ChatGPT tells them there’s no shellfish on the dish, and we tell them there is ,” Madison explains. It’s a phenomenon that’s not just annoying — it can be deadly.
Beyond allergy issues, Madison has seen a growing reliance on generative AI tools for oenology. Customers are rejecting recommendations from restaurant sommeliers , preferring to consult ChatGPT for wine choices. The result? Orders for bottles the restaurant doesn’t have — or that simply don’t exist. AI hallucinations invent wine names, vintages, and producers that never existed, leaving employees to diplomatically handle customer disappointment.
AI Hallucinations in Customer Service: A Growing Phenomenon
The problem isn’t limited to restaurants. Employees at movie theaters, coffee shops , and Apple Stores are reporting similar experiences: customers requesting products or features that don’t exist, based on information provided by an AI chatbot. On Reddit and on X (formerly Twitter), customer service workers are sharing stories of customers treating ChatGPT as an unquestioned authority, refusing to accept corrections from people with real experience and knowledge.
Jason, a federal park ranger in the Mountain West , describes visitors arriving at the park with AI- generated itineraries that are completely absurd — routes that don’t exist, attractions that are closed or inaccessible, distances that have been miscalculated. AI hallucinations in this case aren’t just annoying — they can jeopardize the safety of visitors in remote areas.
The spread of the phenomenon is directly linked to the mass adoption of generative AI tools by the general public. OpenAI ’s ChatGPT , Google ’s Gemini , Microsoft ’s Copilot , and other similar tools have become daily aids for millions of users worldwide. However, this technology has a fundamental problem: language models ( LLMs ) don’t really “know” — they generate text based on statistical patterns, and sometimes that text is just plain wrong.
See also: Airbnb: AI-based customer service launches
Why AI Hallucinations Are So Dangerous for Workers
The problem of AI hallucinations is not new to technology experts, but the mass public exposure to it creates new social dynamics. Customer service workers now find themselves in a paradoxical position: they must refute “information” provided by a technology that the average consumer considers omniscient, without offending the customer. This requires additional emotional labor and diplomatic skills that were not part of their initial training.
Madison describes in detail how she and her colleagues have developed strategies to “mitigate” the impact of AI hallucinations. Instead of directly telling the customer that the AI chatbot gave them incorrect information, they say something like, “We have a very specialized wine list — maybe it shows you a French wine, but we don’t have French wines.” This approach, while effective, is an additional burden for employees who already face multiple pressures at work.
The situation is set to worsen with the release of new AI tools . Muse , Meta ’s new AI agent , is expected to further increase consumers’ reliance on artificial intelligence for decisions that have traditionally relied on human experience and knowledge. AI agents — systems that can perform complex tasks autonomously — bring new possibilities, but also new risks when users trust them without critical thinking.
This phenomenon reflects a deeper social trend: the tendency to attribute authority to technology, even when it contradicts human experience. The old adage “the customer is always right” is being replaced by a new, equally problematic one: “AI is always right.” This mindset undermines the value of professional knowledge and experience, while also creating real risks — particularly in areas such as health, food safety, and navigation in the natural environment.
How to deal with AI Hallucinations as consumers
For consumers, the basic principle is simple: generative AI are useful aids, but not infallible sources of truth. Especially on matters involving health — like allergies — or safety, the opinion of a trained professional should always prevail. LLMs do n’t have access to real-time data about specific restaurants, stores, or locations — and even if they did, they could produce incorrect information with impressive confidence.
For businesses, the phenomenon of AI hallucinations creates new training needs for employees. Employees need tools and strategies to effectively manage situations where customers present incorrect information from AI as indisputable facts. At the same time, businesses could invest in their own specialized AI tools that are based on verified data — for example, a chatbot that knows the exact menu and ingredients of each dish.
See also: Artificial Intelligence: How AI Helps Customer Service
Overall, the phenomenon of AI hallucinations in customer service is a reminder that technology, no matter how impressive, can’t replace human judgment, experience, and accountability. As AI become more prevalent, critical thinking and digital literacy are becoming essential skills for every modern consumer. The next time an AI chatbot gives you information about a restaurant, a product, or a route, it’s worth verifying it with a knowledgeable human — because AI can be wrong, and sometimes that wrongness has consequences.
