Madison: diners trust ChatGPT over servers on allergies

The complaints are not limited to a single restaurant or a single industry. Madison's experience with shellfish-allergy guests who trusted ChatGPT is echoed by a river ranger who watched groups stay skeptical while he pointed at a map, and by an after-school teacher whose students' parents insisted a chatbot's sleep schedule was fine. Lauren Wilford, who has worked in food service for 15 years, describes a shift from guests uploading menu photos to chatbots for wine pairings to agents calling and emailing to make reservations on users' behalf.

A common thread in these accounts is that the chatbot's answer is treated as more credible than the person in front of the customer. Madison and her colleagues have learned to soften the correction, telling guests the wine list is unusually specific rather than saying the bot invented a wine brand. The ranger, Jason, describes the same wall: even with a map and a printed guidebook, some groups remained skeptical.

The stakes here are likely highest where the information concerns safety or care, such as an allergen in a broth or a child's sleep. Whether this pattern eases or deepens may depend on how the next generation of AI agents behaves, since Wilford already suspects an agent, not a person, sent demanding last-minute requests for an event.