New research has found that 90% of AI-generated travel itineraries provide inaccurate information.
British-based travel marketing agency SEO Travel conducted research, asking ChatGPT to plan two-day itineraries for the world’s top-10 city destinations: Paris, Dubai, Madrid, Tokyo, Amsterdam, Berlin, Rome, New York, Barcelona and London.
The results were analysed to check for inaccuracies as well as unrealistic timescales and inefficient use of time. It was found that 90% of the travel itineraries included at least one error.
Over half (52%) of the itineraries suggested visiting attractions, restaurants or cafés outside of their opening hours, and one in four recommended an experience that is permanently or temporarily closed. ChatGPT recommended an 08h00 breakfast at a Tokyo Café, which only opens at 12h00, as well as a trip to the Pergamon Museum in Berlin, which is closed until 2026.
One in four itineraries also lacked logical planning, requiring travellers to either backtrack or make unnecessary detours. Unrealistic schedules were also a recurring problem, such as allocating two hours to watch the Changing of the Guard in London, which takes 45 minutes, but providing a guideline of two-and-a-half hours to visit the Ueno Zoo, Tokyo National Museum and the Toshogu Shrine in Tokyo.
Researchers also found that ChatGPT recommended places that did not exist, such as a fictional eatery in Rome. Finally, the chatbot also tended to recommend exorbitantly-priced options, such as €200 (R3 955) meals, which would be out of the spending range of most travellers.
“Considering there were significant issues and inaccuracies in 90% of ChatGPT-generated travel itineraries, it’s clear that the platform does not provide dependable holiday planning just yet. With this in mind, travel companies remain just as essential and relevant as ever,” said SEO Travel.
