90% plan their trip with AI. Only 2% let it pay

Two figures, placed side by side, explain where artificial intelligence stands in travel in 2026 better than any analysis. Around 90% of consumers say they are interested in planning their trips with generative AI, according to James Waters, Chief Business Officer at Booking.com (March 2026). And only 2% of leisure travellers would let an AI book their trip autonomously, according to data reported by Skift the same year. Between interest and delegation lies an 88-point chasm. That chasm is the story.

The gap, measured from every angle

What stands out is how consistent the pattern is across unrelated sources. Accenture's Consumer Pulse (2026, 25,590 consumers in 16 countries) quantifies it in detail: 89% would use AI agents to plan and search for options, fewer than one in three would hand them the final decision, and only 7% would let them purchase autonomously. McKinsey and Skift reach almost identical proportions by another route: 90% trust AI to get informed about travel; 2% would trust it to book or modify unsupervised.

In Spain the picture is even sharper. Only 4.9% of travellers rank an AI's recommendation as their most trusted source, versus 33.2% who trust their travel agency first (Ávoris Travel Insights, 2026 — a figure from an agency group, worth knowing). And according to the LLYC and Appinio study (April 2026, 2,000 interviews), 28% of Spaniards already plan trips with generative AI, but only 14.9% use it for the final booking. The dominant pattern is "design with AI, verify by hand": among families aged 30 to 44 — the heaviest users, at 31.3% — a third manually double-check what the AI proposes before booking.

Why we don't hand over the card

The first reason is that AI still gets things wrong, and users know it. When Simon-Kucher surveyed more than 10,000 travellers across 10 markets (2025), the complaints were specific: inaccurate answers (52%), generic recommendations (47%), questions it doesn't understand (38%).

The second is structural. Assistants shine at exploring but fail at what Mirai calls the consideration phase: they cannot guarantee verified operational data — is there a cot available? does breakfast have a gluten-free option? what happens if I cancel on Tuesday? That information lives in internal systems, not on indexable pages, and when in doubt the user goes back to Google, the OTA or the phone.

The third is legal: there are still no liability standards for booking errors made by an AI (Skift, March 2026). A trip is not an e-commerce cart: a stay already enjoyed cannot be returned, and nobody wants to find out who answers for a mistake once they are at the airport.

The precedent almost nobody cites

This film has premiered before. For a decade, Google, Tripadvisor and Trivago tried to get users to book without leaving the metasearch engine — so-called "Book on Metasearch". It died quietly: the closure of Tripadvisor Instant Booking on 30 April 2026 was its death certificate. Among the causes documented by Mirai were confusion over who the responsible seller was and users' reluctance to type their card into a "bridge" interface. Agentic booking is, in practice, a second-generation attempt at the same thing, and it inherits exactly those trust barriers. Whoever solves them first — clear seller identity, explicit guarantees, reachable human support — will take the share of the market that is ready to delegate.

Where the gap will close first

Two signals to track. First: corporate travel will adopt agents before leisure does, because it already has the safeguards leisure lacks — company policies, approval flows, defined owners (Skift, March 2026). Second: the gap is not static. Extensive use of AI for planning jumped from 13% to 30% in a single year (Skift Research and McKinsey, 2026). If trust in payment follows a similar curve two or three years behind, today's 2% is the most provisional figure in this entire article.

What I would watch if I sold travel

After fifteen years in tour operator operations, my practical conclusion is that this gap is a window, not a wall. As long as travellers plan with AI but verify by hand, the business that wins is the one that shows up well in the exploration phase (visibility in the assistants) and then makes verification easy: clear data on the website, a reachable human, policies without small print. AI brings the customer to your door; whether they walk in still depends on the oldest thing in the trade — being easy to trust.