Travelers trust AI enough to start. Not enough to book.

Why the AI-referred visitor who spent 70% longer on your site still didn't convert — and where they went instead

The Adobe data from May 2026 showed something that doesn't resolve cleanly. AI-referred visitors to travel sites spent 70% longer per visit than visitors from other sources. They bounced 41% less. By every engagement measure, they arrived with more intent than the average visitor. And they converted 28% less.

High engagement, low conversion. The instinct is to look at the booking path — the rate presentation, the booking engine, the checkout flow. Something in there isn't closing.

That instinct may be looking at the wrong part of the journey.

The verification habit

In March 2026, Yext surveyed 1,120 US adults about how they find and choose local businesses. The findings on verification behavior are the part worth sitting with. Yext sells AI visibility and listings management products, which gives the research a commercial stake, but the methodology is straightforward and the finding is consistent with what other studies are showing.

After receiving an AI recommendation, 62% of consumers immediately search Google for more information. Fifty-eight percent visit the business's website directly. Fifty-two percent click through to sources cited within the AI response. Nearly every AI user verifies — before acting on what the AI told them.

The number that cuts deepest is this one: verification rates are nearly flat across trust levels. High-trust users — consumers who rated their trust in AI recommendations at five out of five — verified via Google at 62%. Neutral users verified at 63%. Trust level doesn't predict whether someone verifies. It only predicts how.

The AI recommendation is not a closing signal. It is an opening one.

Where the loop runs

The verification behavior follows a predictable sequence: AI recommendation, then Google, then the business's own website, then sources the AI cited. For a hotel, that sequence has a specific shape.

Google returns OTA listings, review aggregators, metasearch results, and the hotel's own organic presence — roughly in that order for most independent properties. The traveler who arrived at a hotel's direct site from an AI recommendation and didn't convert didn't necessarily lose interest. They went to verify. And verification ran them through Booking.com, TripAdvisor, and Google Hotels before it brought them back — if it brought them back at all.

This is what the conversion gap is measuring. Not a failure of the booking path, but a failure to hold position through a verification loop that routes through intermediaries by default. The AI referral was the start. The OTA was the close.

The trust architecture hotels are navigating

The Fractl and Search Engine Land study from Q2 2026 established that consumer trust in AI search recommendations fell from 82% to 54% in a single year, while usage kept rising. The gap between those two figures is the verification loop — consumers using AI more but trusting it less, filling the gap between the AI answer and their purchase decision with additional sources.

For high-stakes purchases, that behavior intensifies. Hotel stays sit in the category where the cost of a wrong recommendation is real: a bad hotel is two nights of a ruined trip. The higher the personal cost of being wrong, the more sources a consumer checks before committing. AI gets them started. The verification layer is where the decision actually forms. That's not new consumer behavior. It's the same behavior that made TripAdvisor indispensable fifteen years ago, running on newer infrastructure.

Hotels have spent a decade optimizing for Google's verification layer — building review volume on TripAdvisor, maintaining rate parity across OTA listings, managing their Google Business Profile. That work didn't disappear when AI arrived. It became the verification infrastructure that AI-referred travelers land in. Where that infrastructure routes the traveler depends on what it was built for and what it returns when checked.

What the engagement data actually shows

The 70% longer session and 41% lower bounce rate aren't signs of confused visitors. They're signs of visitors doing what Yext's data says consumers do: verifying. An AI-referred traveler who spends 70% longer on a hotel's direct site is reading room descriptions, checking availability, looking at photos. They're building a case. Whether that case closes on the direct site or migrates to an OTA depends on what they find when they look — and what they find when they leave to check.

The conversion gap has narrowed sharply since late 2024, from 86% to 28% in Adobe's tracking. The direction is consistent. What's driving the narrowing isn't visible in the data Adobe published.

The 28% that remains is the portion of the loop still terminating somewhere else.

The implication for hotel commerce

AI visibility investment gets travelers into the consideration set. The verification loop is where they decide. Those are two different problems, and they require two different kinds of work.

The first is an eligibility problem — being present and accurately represented in AI responses. The second is a retention problem — being strong enough across the verification surfaces that the journey completes on the hotel's own terms rather than an intermediary's.

Neither is new. Both have been the work of hotel distribution for twenty years. What AI changes is the sequence: the referral now arrives before the verification, rather than after the search. The loop is the same. The entry point moved.

by Markus Busch, Editor and Publisher of Hospitality.today

Read also: AI can find your hotel. It just won't recommend it.

Enjoying this analysis? Hospitality.today delivers daily insights on hotel distribution, AI trends, and travel commerce — straight to your inbox. Subscribe for free at Hospitality.today →