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 AI travel users 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 Survey 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 moving rapidly into the travel-planning mainstream. Nearly half of travelers surveyed said they had used or tried AI for travel, but just 11% had used it both to plan a trip and book something.
That gap points to the next challenge for AI in travel: not simply helping travelers decide what to do, but ensuring those answers hold up in the real world.
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. Nearly half of U.S. leisure travelers surveyed, 49%, said they had used or tried AI for travel.
So far, however, AI appears to have made greater inroads in planning than booking. 33% had used AI to research or plan a trip but not to book anything, while 11% had used it both to plan and to book. Another 5% had tried AI for travel but didn’t find it useful.
And as more travelers turn to AI for advice, the reliability of that advice becomes increasingly important.
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:
Respondents could report more than one problem.
Perhaps the most striking finding: roughly one in six AI travel users 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.
Those experiences appear to matter for how much travelers trust AI afterward.
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.
The survey does not establish that a problem with a recommendation 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. And once someone starts checking an AI-generated itinerary against Google, Reddit, online reviews, a hotel concierge, or a local expert, some of AI's promise disappears.
Instead of eliminating research, AI risks becoming another source travelers feel they need to vet.
That hesitation becomes even more consequential when AI moves from recommending something to actually acting on a traveler’s behalf.
Greetwell asked travelers how much they would trust an AI tool to choose and book something on their behalf without double-checking it themselves.
The majority expressed little or no trust across several major travel purchases: 68% had little or no trust in AI choosing and booking a flight without them double-checking it. Majorities said the same for a rental car (65%), hotel (64%), local tour or activity (63%), and even a restaurant reservation (54%).
The pattern points to an important distinction. Travelers may be increasingly comfortable asking AI what they should do, but they're considerably less comfortable letting AI do it for them.
When travelers are asked whom they would actually turn to for advice, AI still trails more traditional sources. We also asked all 1,000 travelers a simple question:
Imagine you have one free afternoon in a city you don't know well. Whose recommendation would you take?
AI is extraordinarily well suited to this kind of question. Yet when travelers were forced 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 something goes wrong. Greetwell asked travelers who they would want to handle a last-minute cancellation or refund for a booked tour or activity. 65% said they would want a real person they could call or message. Just 5% preferred an AI assistant or chatbot.
The issue may not be AI’s ability to generate an answer. It’s confidence that the answer reflects the real world, and that someone will be there when the real world doesn’t go according to plan.
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 or incomplete 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.
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 article use the respondent-level survey weights supplied with the survey data. Percentages may not sum to 100% due to rounding. Respondents could select more than one problem, so individual percentages should not be summed.
Of the 1,000 respondents, 485 reported having used or tried AI for travel. These respondents form the base for findings specifically concerning experiences with AI travel recommendations.
Comparisons between subgroups describe associations within the survey sample. They should not be interpreted as demonstrating that one experience caused another.