AI hallucinations are making entitled customers even worse

Jason, a river ranger who works on federal lands in the Mountain West, says he’s noticed an increase in visitors arriving to the park with nonsensical AI-generated itineraries.

“My colleagues and I have explained that the AI hallucinated campsites or scheduled things out of geographical order, so the group would have to row their boats upstream some days (not possible) and float with the river current other days,” says Jason, who asked to be referred to by a pseudonym because he’s a federal employee. “The craziest part is that even with a ranger explaining this while pointing at a map, we’ve had multiple groups remain skeptical. A certain type of AI user is inclined to trust ChatGPT over rangers, maps, and printed guidebooks.”

Casey, an after-school program teacher in New York who asked to be referred to by a pseudonym, says she’s noticed many students’ parents, nannies, and other caretakers are now turning to ChatGPT or Claude for their questions on childhood development.

“We had one family that consulted ChatGPT to come up with a bedtime routine and sleep schedule for their kid, but their kid wasn’t getting quite enough sleep,” Casey says. Based on his behavior in the program, Casey estimated that the child needed two to three more hours of sleep a night, and told his parents this. “The parents just insisted that their ChatGPT was telling them that their schedule was fine.”

Some parents, she says, now see generative AI tools as experts whose knowledge surpasses that of childcare professionals. “It can be really hard to talk with somebody who has made their mind up,” Casey says, “because they trust this outside entity as being invaluable.” She’s also noticed more anodyne uses of the technology, like kids coming in with AI-generated coloring sheets.

Lauren Wilford, who has worked in food service for 15 years, has noticed AI creeping into her recent interactions with guests as well. It started with the menu: Someone would upload a photo of it to ChatGPT or Claude and ask it what to order, or for guidance on a wine pairing, instead of asking the server.

But now, it appears, some customers have dropped the intermediary altogether. Over the past year, Wilford’s restaurant has been getting calls and emails from agents trying to make reservations on users’ behalf. The rise of agentic AI makes these interactions that much more confusing.

Wilford was surprised when a hands-off client suddenly became demanding on the eve of a big event, making last-minute requests for podiums and AV equipment — neither of which were readily available at the Providence, Rhode Island, restaurant she manages — and eleventh-hour changes to the catering menu. But Wilford was even more taken aback when the clients arrived the next day and didn’t mention any of it. Though she can’t prove it, Wilford believes the requests may not have come from anyone at the company at all; she suspects they were the work of an AI agent.

“I’m sympathetic to the reasons people use AI agents,” Wilford says, especially for bureaucratic and administrative tasks. “But anything where there’s value to two humans seeing each other and working something out together, I think we’re not sensitive to what might be lost if we try to skip all of that stuff.”