AI plans holidays, but fails on emotions

Artificial intelligence (AI) is rapidly becoming a travel agent for millions of holidaymakers, but most travellers are drawing the line at allowing it to spend their money.

New research shows that AI is almost as popular as Google for starting holiday planning, yet only 5% of travellers book exactly what AI recommends. This suggests that travellers are embracing the technology for inspiration while remaining deeply sceptical about its decisions.

The findings come from a September 2026 study by US marketing agency Mower, which surveyed 1,000 American leisure travellers aged 18 to 64 who use AI at least occasionally. It found that 28% would be most likely to begin planning their next holiday with AI, compared with 30% who would start with Google.

Once AI has made its recommendations, however, 59% turn to Google to investigate further, while 49% visit the official websites of hotels, airlines or destinations. Another 40% compare prices across multiple sites.

Jarrod Lopiccolo, co-founder and CEO of US digital marketing and experience agency Noble Studios, believes the figures reveal a fundamental limitation of how AI approaches travel.

“I think AI struggles with the emotional side of travel. It can recommend the perfect itinerary based on everything it knows about you, but sometimes the best travel experiences are the ones you never knew you wanted.”

That is a significant limitation for a technology increasingly being promoted as capable of organising an entire holiday, from selecting destinations to booking flights and accommodation.

AI can assemble itineraries based on budgets, dates, interests and previous preferences. It can recommend restaurants according to dietary requirements, suggest attractions within walking distance of a hotel and reorganise a day’s sightseeing around the weather forecast.

But an itinerary optimised around a traveller’s known preferences may also exclude the surprises that make a holiday truly memorable. Personally, the moment this struck me most forcibly was experiencing street food in Mexico City, from a street vendor AI knew nothing about. It all but made my exploration of the city.

On the other hand, AI thinks a traveller who routinely chooses international hotel groups will keep wanting familiar brands, overlooking an independent establishment with a distinctive local character.

“I actually think independent hotels have a tremendous opportunity here,” says Lopiccolo. “Their personality, authentic guest experiences and local connections can help them stand out, provided they’re doing a great job of making that information accessible to AI.”

Noble Studios works with travel destinations on precisely this challenge. As travellers increasingly ask AI where to go and what to experience, destination marketers are having to reconsider how their information reaches potential visitors.

“We’re helping destinations rethink how they tell their stories in a world where AI is increasingly influencing travel decisions. That means making sure their unique experiences and authentic stories are discoverable, credible and ultimately recommended by these platforms.”

Clear descriptions of accommodation, facilities and local attractions become potential inputs into recommendations produced by ChatGPT and similar services. But the business still has to persuade that traveller to book.

The Mower findings reveal how substantial that second hurdle remains.

“I think they’re looking for reassurance,” said Lopiccolo of the 59% who return to Google after consulting AI. “AI might give you a great recommendation, but when you’re about to spend thousands of dollars on a vacation, you want to know it’s real, the reviews back it up and you’re making a good decision.”

The research offers revealing distinctions between different kinds of travel.

Among the AI-using travellers surveyed, only 28% mostly or completely trusted AI to book a flight on their behalf, while 29% felt the same about booking a hotel. Almost half were comfortable relying heavily on AI to plan a road trip, but that figure dropped to 14% for a luxury holiday costing $5,000 or more.

For a honeymoon, it fell to just 11%.

Obviously, the consequences of a disappointing restaurant recommendation are relatively minor. An incorrectly booked international flight, particularly one involving multiple airlines or connections, becomes an expensive problem.

It also raises a question that travel technology companies will have to confront as AI agents progress from recommending holidays to completing bookings and payments: Who pays when the AI gets it wrong?

“If an AI platform is going to take responsibility for booking someone’s vacation, it needs to take responsibility when things go wrong, too,” says Lopiccolo. “Travelers shouldn’t be left holding the bag because an algorithm made a mistake.”

The question becomes especially complicated when an AI agent interacts with several independent booking systems. If it selects the wrong date or misunderstands a cancellation condition, the airline or hotel may have fulfilled the reservation it received perfectly correctly.

The traveller could then discover that responsibility for the error lies somewhere between the AI provider, the booking intermediary and the service provider. Lopiccolo’s position is a proposed principle of accountability, although contractual terms and consumer protection laws will determine liability in individual cases.

The reluctance to surrender control extends well beyond the American travellers surveyed by Mower.

Booking.com’s 2025 Global AI Sentiment Report, based on responses from 37,325 consumers in 33 markets, found that 89% wanted to use AI in future travel planning. Yet only 12% were comfortable allowing AI to make decisions independently.

The research found that 65% expected autonomous trip planning to become mainstream in the near future. In other words, consumers anticipate a future in which AI organises travel independently, while remaining reluctant to hand over that authority themselves.

Even Lopiccolo, whose business helps tourism brands reach travellers through AI-driven discovery, draws a distinction.

“I’d happily let AI do most of the research and planning, but I’m not quite ready to hand over my credit card and let it make every decision. I actually enjoy some of the planning process, and I want to keep the freedom to change my mind along the way.”

Lopiccolo believes the surprisingly low proportion of people booking exactly what AI recommends reflects something deeper than doubts about the technology’s accuracy.

“Because travel is personal, and people still want to feel like they’re making their own decisions. I see AI becoming an incredible travel companion, but I don’t believe most people are ready to give up control of their entire experience.”

* Arthur Goldstuck is CEO of World Wide Worx, editor-in-chief of Gadget.co.za, editor of GadgetWings, and author of “The Hitchhiker’s Guide to AI – The African Edge”.