The AI Travel Report 2026
AI can plan a trip in seconds. Convincing travelers they can trust the result may be the harder challenge.
AI can recommend virtually anything to a traveler, including, apparently, things that don't exist. Roughly one in six people who had used or tried AI for travel said they'd been recommended a place, tour, or activity that wasn't real.
That's one of the more striking findings from the Greetwell AI Travel Report 2026, an August survey of 1,000 U.S. leisure travelers. Among the 485 respondents who had used or tried AI for travel, 55% encountered at least one recommendation that was inaccurate, unavailable, nonexistent, closed, or otherwise wrong.
The findings come as AI is quickly becoming part of travel planning. More than half of travelers surveyed had used or tried AI for travel, while 14% had used it both to research or plan a trip and book something.
As AI becomes a bigger part of travel, the challenge may be shifting from generating good answers to ensuring those answers hold up in the real world.
Key findings
14% have used it both to research or plan a trip and to book something.
Inaccurate, unavailable, nonexistent, closed, or otherwise wrong. 16% were recommended something that didn't exist.
68% would have little or no trust in AI choosing and booking a flight without double-checking it; 63% say the same about a local tour or activity.
Asked to pick one source, travelers chose friends and family (31%) and hotel or local staff (25%) well ahead of AI.
AI is quickly becoming part of how people plan travel
AI is moving rapidly into the travel-planning mainstream. Deloitte's 2026 summer travel research found that GenAI use for trip planning rose to 25% of travelers, up from 15% a year earlier.
Greetwell's survey offers a snapshot of how far that experimentation has spread. More than half of U.S. leisure travelers surveyed, 53%, had used or tried AI for travel.
So far, however, AI appears to have made greater inroads in planning than booking. 34% had used AI to research or plan a trip but not to book anything, while 14% had used it both to plan and to book. Another 5% had tried AI for travel but didn't find it useful.
Travelers are increasingly willing to ask AI where to go.
Far fewer have crossed the line from asking AI for advice to using it to make a booking.
What goes wrong with AI travel recommendations
Among travelers in the survey who had used or tried AI for travel, 55% encountered at least one recommendation that was inaccurate, unavailable, nonexistent, closed, or otherwise wrong.
Perhaps the most striking finding: roughly one in six people who had used or tried AI for travel said they had been recommended a place or activity that didn't exist at all.
That's a particularly consequential type of AI mistake. Vacation time is finite. If AI sends you across the city for an experience that doesn't exist, or to an attraction that closed hours ago, you can't get that time back.
Bad recommendations are associated with declining trust
Overall, trust is moving in a positive direction among travelers who have used AI for travel. 34% said their trust in AI travel recommendations had increased over the previous year, while 14% said it had declined.
But travelers' experiences with AI appear closely associated with that trust. Among travelers who had not encountered a problem, just 5% said their trust had declined. Among those who had encountered at least one problem, 21% said so — more than four times the rate.
The survey does not establish that a problem caused someone's trust to decline, but the difference between the two groups is substantial.
Every inaccurate recommendation gives a traveler a reason to verify the next one. Instead of eliminating research, AI risks becoming another source travelers feel they need to vet.
Travelers aren't ready to hand AI the keys
Greetwell asked travelers how much they would trust an AI tool to choose and book something on their behalf without double-checking it themselves. Majorities expressed little or no trust across every major travel purchase — and the pattern points to an important distinction: travelers may be comfortable asking AI what they should do, but they're considerably less comfortable letting AI do it for them.
When it really matters, travelers still turn to people
“Imagine you have one free afternoon in a city you don't know well. Whose recommendation would you take?”
This is precisely the kind of open-ended travel question AI tools are designed to answer. Yet when all 1,000 travelers were asked to choose one source, friends and family, hotel or local staff, and online reviews all ranked well ahead of AI.
The preference for the human touch becomes even more pronounced when a plan falls through. Presented with a scenario in which a booked tour or activity is cancelled the day before because of bad weather, and asked who they would want handling the rebooking or refund, 65% said a real person they could call or message. Just 5% preferred an AI assistant or chatbot.
The answer isn't humans versus AI
These findings don't suggest that AI has no place in travel. In many ways, travel is exactly the kind of problem AI is well suited to solve.
Planning a trip remains remarkably fragmented. Travelers move between search engines, maps, review sites, social media, hotel websites, booking platforms, and recommendations from friends just to decide where to go and what to do. AI can bring much of that process into a single conversation. But there's an important difference between generating a convincing recommendation and delivering a reliable one.
Travel information is unusually dynamic. Hours change. Tours sell out. Availability disappears. A recommendation can be accurate one week and useless the next. That creates a particular challenge for AI systems that can present outdated information with the same confidence as current information.
The next frontier for AI in travel may be less about generating more recommendations and more about connecting what travelers want with reliable information about what's actually happening on the ground.
AI has already proven it can build an itinerary in seconds. The harder challenge is making sure that itinerary still exists when the traveler gets there.
Charts for republication
Free to use with attribution to the Greetwell AI Travel Survey 2026
Methodology
Greetwell commissioned an online survey of 1,000 U.S. adults who had taken at least one leisure or vacation trip involving an overnight stay during the previous 12 months. The survey was conducted in August 2026.
All percentages reported in this report use the respondent-level survey weights supplied with the survey data. Percentages may not sum to 100% due to rounding. For questions where respondents could select more than one answer, individual percentages should not be summed.
Of the 1,000 respondents, 485 had used or tried AI for travel. After weighting, this group represents approximately 53% of the survey sample. Findings specifically concerning experiences with AI travel recommendations use these 485 respondents as their base.
Comparisons between subgroups describe associations within the survey sample and should not be interpreted as evidence that one experience caused another.
Press contact
[email protected]Interview requests, additional cross-tabs, and custom cuts of the data are available on request. Please attribute findings to the Greetwell AI Travel Survey 2026.