In the past twelve months, 56% of US leisure travelers used AI to plan, book, or get in-destination help on at least one trip, according to Phocuswright's AI Surge research released in March 2026. That number was 33% in the first half of 2025 and 43% by the end of last year. Trip planning with AI is now the default behavior, not the experimental one.
The booking is not.
Expedia Group's AI Trust Gap survey of 5,700 adults across the US, UK, and India, released April 14 2026, found that 53% of travelers are comfortable letting AI suggest travel options. 66% would not trust an AI assistant to make a booking on their behalf. Same person. Same trip. Two completely different levels of trust depending on which step of the funnel the AI is operating in.
We built Concier around this exact split.
The trust gap between suggestion and booking
Travelers are eager to let AI search, summarize, and recommend, but they pull back when the action commits real money or real plans. Expedia's April 2026 survey found 53% comfortable with AI suggesting travel options against 66% who would not trust AI to make the booking, a 13-point cliff in a single survey. 68% prefer to book with a trusted brand over an AI chatbot or agent.
The gap is wider than the headline. The same Expedia study found 57% of travelers are concerned about loss of control, data privacy, and payment privacy in AI-driven transactions. These are not soft signals. They are the reasons people are not pressing the button.
What's interesting is how happily the same travelers use AI earlier in the journey. Phocuswright reports that the share of US travelers using generative AI platforms for trip research has risen fivefold since 2024 and now sits within striking distance of traditional search engines for that same task. The same humans who let ChatGPT or Gemini build a four-day itinerary in Lisbon will not let it press confirm on the hotel.
Why travelers approve before they book
The trust gap is not about AI competence. It is about consequence. A bad recommendation costs ten minutes of scrolling. A bad booking costs money, a vacation, sometimes a relationship. Travelers want to verify, and they want to verify on a brand they already know.
Mike Coletta, senior manager of research and innovation at Phocuswright, put it directly:
Half of travelers who used AI in search engines told us they still clicked through to source websites after seeing AI answers in search. This violates the common narrative of a zero click world. AI is definitely reducing clickthrough in search overall, but travel is much more resilient because it's higher stakes and verification-heavy, especially in the transaction phase.
That sentence describes the entire shape of the next funnel. AI does the early work. The human still clicks through. The booking still gets verified on a brand surface. The handoff is the product.
The trust gap is also where the value lives. Phocuswright's May follow-up found that AI travelers have a median household income of $129,200 against $104,000 for nonusers, take 3.8 leisure trips a year versus 2.9, and spend $4,500 annually on leisure travel against $3,000. The travelers most fluent with AI are also the ones spending the most. The funnel that captures them is the funnel that captures the next decade of leisure-travel revenue.
What "AI plans, human approves" actually looks like
The pattern is a two-step interaction. AI does the research, synthesizes the options, and presents a short list. The traveler reviews, asks follow-up questions, and confirms. The booking happens with the traveler's explicit sign-off, on a brand surface the traveler already trusts. The AI is in the loop, not in the driver's seat at the cash register.
In practice it looks like:
- A traveler asks an AI assistant for a long weekend somewhere in northern California, two adults, dog-friendly, walkable, under $500 a night, no chains. The assistant builds three options, each with hotel, route, and a couple of restaurants.
- The traveler reviews the short list, asks follow-up questions ("which one has the better breakfast?"), and picks one.
- The booking happens on the hotel's direct site, not inside the assistant. The traveler sees the confirmation in their own inbox.
Note where the trust moves. The assistant is trusted with discovery and curation. The brand is trusted with the transaction. The handoff between them is where most travel software still breaks, and where Concier sits.
What this means for hotels chasing AI traffic
The trust gap is not bad news for hotels. It means the AI referral wave is real but it stays funneled through a human moment. Hotels that show up in the AI's short list still get to win the booking on their direct site. The job is two-fold: be discoverable at the AI layer, and convert when the human lands on the brand surface to approve the booking.
The first half of that job is what we covered in Boutique Hotels and the Coming AI Discoverability Layer. Independent hotels that do not show up in the assistant's short list are functionally invisible to the 56% of travelers who now plan with AI. BCG, in its March 2026 AI-First Hotels report, frames the same problem more sharply: a property without a coherent digital footprint risks becoming a "digital ghost, absent from the top three recommendations."
The second half is the on-site experience. A clear direct-booking flow. Reasons to believe stacked in copy and in structured data. A price that does not lose to the OTA listing one screen below. Hotels that nail both halves capture the booking. Hotels that nail only the first generate the referral and lose the conversion to whatever the traveler trusts more. Hotels that nail neither become OTA inventory.
The funnel that's emerging
The old funnel (search, click, click, click, click, book) is being replaced by a conversational front end and a confirmation moment. AI does the planning in dialogue with the traveler. The human approves on a brand site or a trusted booking surface. The middle five clicks are gone. Trust gates the last step, not the first.
We wrote about why this funnel was collapsing in The Booking Funnel is Collapsing into Conversation. The new Phocuswright and Expedia data is the proof. Travelers are already living in the conversational front end, and they are explicitly pulling the booking moment back into a trusted brand experience. They are not abdicating the decision. They are upgrading their research.
For travelers, this is a relief. The browser-tab tax (the forty minutes of comparing across five sites) drops. The control over the final commit stays. For hotels, the implication is sharper. The competition is not who ranks first on Google any more. It is who gets cited by the AI assistant and converts the human two clicks later.
We are building Concier for that exact handoff. If you travel and you have thoughts on what the AI-plans, human-approves flow should feel like, join the waitlist. The product is being built around the people who already live in the trust gap, not around the people who haven't found it yet.