Don’t trust ChatGPT to make your travel plans, yet

Article content

Trusting your vacation planning to artificial intelligence (AI), may not be so clever, a survey suggests. A study found a resounding 90 per cent of travel itineraries generated by ChatGPT, a bot that answers questions in conversational language, to be strewn with errors.

Article content

The blunders included listing shuttered restaurants and attractions, as well as wrong opening times and impossible-to-meet schedules.
The research was conducted by digital marketing firm SEO Travel, which tasked ChatGPT with creating two-day itineraries for London, New York, Paris, Tokyo, Madrid, Barcelona, Berlin, Amsterdam, Dubai and Rome. But researchers found the results “riddled with inaccuracies,” with 90 per cent containing one or more errors.

Article content

One in four (25 per cent), of the holiday plans recommended a site that was temporarily or permanently closed, and more than half (52 per cent), suggested visiting at least one attraction outside its opening hours. Almost one-third (30 per cent), included a Michelin-starred restaurant, despite their traditionally steep prices.

Article content

One in four of the itineraries included needless diversions — for instance, a 20-km detour for breakfast in Dubai. Some also recommended visiting non-existent places.

SEO travel director Tom McLoughlin told Travel Weekly that as many as two-thirds of travellers expect to use AI in holiday planning.
But he warned: “Don’t fall into the trap of thinking AI is the answer to everything. AI will always be feeding off existing information, so it is always going to be behind or feature inaccuracies.”

He said bigger travel firms will have the resources to build their own AI tools, which will likely be more accurate than bots now on the market.
“Smaller companies need to be more vigilant,” he said.

Article content