Only 5% of consumers act on an AI recommendation without checking anything first, and 49% go straight to the business's own website — a dynamic that gives independent hotels a commission-free discovery channel that closes the moment AI agents take the payment.

Yext surveyed 3,848 adults across the US, UK, France, Germany and Italy this spring on what they do after an AI recommends a local business. Beyond the 5% who acted without verification, 53% searched Google or Bing. Trust barely moved the number: US respondents who rated their trust in AI five out of five checked a search engine at 62%, versus 63% for neutral users.

The pattern holds for travel specifically. Adobe's traffic data shows visits from AI tools to US travel sites rose 119% year over year in July. Those visitors were 21% more engaged than other traffic and converted only 1% less — a year earlier the conversion gap stood at 47%. They arrive primed to read the page closely.

What replaces the billboard effect?

The mechanics echo Cornell's older research. In 2009, Chris Anderson of Cornell's Center for Hospitality Research listed four JHM hotels on Expedia, removed them, then relisted them. During the visible weeks, reservations through the hotels' own channels rose 7.5% to 26%, on top of Expedia's own sales — the "billboard effect." When Anderson and Saram Han revisited it in 2017, the effect had weakened but persisted, because travelers still looked at an OTA before booking.

That look now starts with an AI answer. A traveler asks ChatGPT for a quiet seaside hotel, gets three names, and checks them. One of the places she checks is the hotel's own site — the only page in her search where no competitor sits beside it. The industry calls the resulting direct booking the Shortlist Effect. Unlike the OTA billboard, it carries no bill at all.

Why the website decides the commission

The economics turn on conversion. If the hotel's page fails to convince her, she returns to Booking.com and books the same room she viewed ten minutes earlier — now carrying a commission that starts at 15% and climbs for hotels in Booking.com's visibility programs. The AI made the introduction for free; the hotel pays for the close.

Independents face a sharper test than brands. A Courtyard guest knows what she gets before she clicks; the brand has done the checking for her. The independent has only the AI's one-line description — sea view, good breakfast, quiet at night — and its website must prove that sentence true and price it sensibly.

How long does the window stay open?

Lighthouse, the hotel software vendor, put AI tools at just under 1% of visits to hotel websites in July, after a jump of more than 50% within weeks. The visit survives because paying inside an AI remains rare and distrusted — and that is being engineered away. In late August, Google switched on hotel booking inside AI Mode in the US, with Google Pay at checkout. Every step that lets the traveler pay inside the answer makes the hotel website optional, and the commission goes to whoever holds her card.

Until then, the operational advice is concrete: ask ChatGPT for a hotel in your market with dates, a budget and one priority; if the property appears, compare the AI's sentence against the site's first screen, then place that screen beside the same dates on Booking.com. Whichever makes the case faster wins the booking.

Can anyone measure it?

Not yet. Anderson needed Expedia's cooperation to toggle listings on and off; no one can switch off ChatGPT, and no property can attribute these visits to bookings. The Yext and Adobe figures come from vendors with commercial stakes and cover categories beyond hotels, and Booking.com commissions vary by contract. The Shortlist Effect remains real but uncounted — and dated by the arrival of in-agent payments.