AI travel advice can send people off course, from non-existent sites to visa mix-ups

The Story

A roundup of AI travel-planning errors warns that automated travel advice can be wrong, sometimes with serious consequences.

One example described hikers planning a trek based on an AI suggestion for the “Sacred Canyon of Humantay,” which was later said to not exist; the would-be destination was described as being at high elevation where the warning became crucial.

Another example involved visa guidance for entering Chile, where an individual reportedly relied on ChatGPT’s “no visa required” assurance and then was unable to board a flight to Chile, with the issue escalated enough that Australia’s foreign affairs department publicized the mistake and warned travelers not to blindly follow AI responses.

A separate account described a Singaporean traveler using Notion AI to generate a 12-day itinerary for Vietnam, but the plan included a day trip to Ba Na Hills while the traveller was based in Hanoi, a trip described as 1,400 kilometres there and back.

The roundup also cited a magic-themed Lisbon recommendation in which ChatGPT pointed to the “Palácio do Mago,” which was later described as a hallucination after the chatbot was pressed for details.

Additional examples included AI suggesting the Eiffel Tower was in Beijing and recommending a restaurant as suitable for infants that, the roundup said, had later been shown to have “relocated” to Greenland.

The compilation also referred to airline-related fallout, including a case involving Air Canada compensation after an AI chatbot told a customer he was eligible for a refund for a discounted bereavement fare that the roundup said was not actually eligible.

The roundup said travel planning problems arise when chatbot information is out of date or when the system fills gaps with fabricated details, and it advised travellers to independently check itinerary details, including visas and booking terms.