When two tourists in Peru pulled out their phones to show Miguel Angel Gongora Meza their hiking plans, the seasoned guide couldn’t believe what he saw. They were bubbling with excitement about trekking to the “Sacred Canyon of Humantay.”
The reason why Gongora Meza was so flabbergasted is that no such place exists.
“They [showed] me the screenshot, confidently written and full of vivid adjectives, [but] it was not true,” Gongora Meza told the BBC. “There is no Sacred Canyon of Humantay! The name is a combination of two places that have no relation to the description.” The pair had already spent nearly $160 to get dropped off on a rural road with no guide and no actual destination.
This could have gone horribly wrong. In Peru’s Andes, mistakes like this can be dangerous. “This sort of misinformation is perilous in Peru,” Gongora Meza warned. At 4,000 meters up, where there’s no phone signal and little oxygen, an imaginary destination can become someone’s final stop.
When AI Dreams Up Places That Don’t Exist
Generative AI tools like ChatGPT, Microsoft Copilot, and Google Gemini have become travel companions for millions. A recent survey found that nearly a third of international travelers now use them. Dedicated travel apps like Wonderplan and Layla promise perfectly optimized itineraries, as if your vacation could be engineered like a physics problem.
But these AI systems don’t actually understand geography or logistics in the same way that a person would. They generate words that sound plausible. That means your “romantic mountain hike” can turn into a sunset stranding, like what happened to traveler Dana Yao in Japan.
ChatGPT told her the last ropeway down from Mount Misen left at 17:30. In reality, it had already closed. “So, we were stuck at the mountain top,” Yao told the BBC.
Sometimes the hallucinations are laughable. Layla once told users there was an Eiffel Tower in Beijing. Other times, the errors come wrapped in logistical nightmares: attractions suggested just before closing time, “short walks” that turn out to be miles long, or, in one case reported by the Huffington Post, an Alaska itinerary that skipped a helicopter ride entirely (that was listed as a non-negotiable requirement in the original prompt) and scheduled driving immediately after 24 hours with no sleep.
Even when AI sticks to real places, it often ignores the messy realities of how people actually travel. Judy Gauthier of Go City recalled how AI sent her family to Laurel Falls in the Smoky Mountains — only for her to later discover via Facebook that the trail had been closed for 18 months.
When she asked about swimming in San Sebastián, Spain, AI warned her of shark attacks. The machine had confused the city’s aquarium with the actual beach.
Why AI Trips Up on Travel
The failures come from how large language models work. “It doesn’t know the difference between travel advice, directions or recipes,” explained Rayid Ghani, a machine learning professor at Carnegie Mellon University, to the BBC. “It just knows words. So, it keeps spitting out words that make whatever it’s telling you sound realistic.”
Travel logistics are especially tough. Transportation schedules change, weather shifts, and local quirks matter. Dr. Niusha Shafiabady, a computational intelligence researcher at Australian Catholic University, said that “flight availability, prices, weather and traffic shift in real time, making it hard for AI to stay updated.”
Even with complete information, AI bombs at travel benchmarks. Recent research shows OpenAI’s most advanced model succeeds only 10% of the time at complex trip planning.
Jay Stevens, CEO of Wayfairer Travel, put ChatGPT to the test for a ski trip in Japan. It suggested spectacular onsens (Japanese hot springs) but left him stranded when bus schedules didn’t match reality. “So, at 11 minutes past 11, it’s snowing, so I’m covered in snow and starting to feel a bit cold,” he told Travel Weekly. He eventually caught a taxi, but not before realizing the stakes of trusting an algorithm with rural logistics.
“Once you get outside of big data, you’re into smaller data, it can find whatever it wants to find and start believing in it,which it clearly does,” Stevens said.
Real travel advisors factor in quirks, such as airport security times, local holidays, and the fact that families don’t want to zigzag across a city. As travel consultant Nolan Burris told HuffPost: “While AI is great for exploring ideas, using it for actual travel planning is a gamble.”
So, Should You Trust AI With Your Vacation?
AI-generated itineraries work best as brainstorming tools, not blueprints. Experts recommend verifying every suggestion — check hours on official websites, confirm closures on local Facebook groups, and cross-reference distances on Google Maps. Build in buffer time.
Or think of it this way: AI can’t bend the laws of physics. It might tell you to end a museum tour at the exact time your plane departs. Until teleportation is real, don’t expect your AI assistant to get you from gallery to gate in zero minutes.
In the end, these digital planners are closer to overconfident interns than expert travel guides. They can hand you a list of great ideas, but you’ll need to sort the fantasy from the feasible.
And if you do happen to get stranded after trusting some AI’s travel advice, might as well make the best of it. “Try to shift the disappointment [away from] being cheated by someone,” said psychotherapist Javier Labourt. “If you are there, how will you turn this [around]? You’re already on a cool trip, you know?”
