AI Agents Drive Under 2% of Bookings — But 2027 Commerce Terms Are Being Set Now

Large language models account for well under 1% of Booking Holdings' room nights, and AI referrals sit below 2% at both Booking and Marriott, according to Bernstein analyst Richard Clarke. Yet the commercial terms governing AI-driven booking are being written this year, and sellers who wait will inherit pricing set by someone else's integration.

The gap between negligible volume and hardening terms defines every travel commerce budget heading into 2027. Four positions are moving now.

Who Is the Merchant When an Agent Books?

Meta's Muse agent books flights through Duffel's direct airline connection, with loyalty numbers and cancellations supported. Asked for a New York hotel room, the same agent opened its own browser and chose between Expedia and Hotels.com unprompted. The airlines in that search have a defined commercial relationship with the agent. The hotels were shopped by software they never signed with.

Building direct AI capability has limits too. Google's agentic hotel booking in AI Mode keeps the hotel or OTA as merchant of record, but passes along only name, payment, occupancy, and property details. The conversation that produced the booking stays with Google. An integration buys a place in the transaction — priced as distribution access, with little customer data attached.

Does the Payment Rail Decide Loyalty Economics?

Muse pays airlines with a one-time virtual card through Link by Stripe. Google's tool settles through Google Pay. Hilton's chief information officer said consumers are already handing agents their card details, and Booking Holdings CEO Glenn Fogel told the Skift Global Forum that trust around payment details decides who gets the booking.

The stakes are highest where loyalty economics run on card spend. Delta collected $8.2 billion from American Express in 2025, about 14% of adjusted operating revenue. Every agent checkout that defaults to its own wallet influences which card funds the trip.

Can Loyalty Survive Algorithmic Price Comparison?

Hilton expects agents to reshop bookings until departure, and travelers already do: half of 1,000 U.S. travelers in a McKinsey and Skift Research study changed a confirmed stay after spotting a better option. Skift Research found 44% of U.S. travelers book outside their preferred program on price, and 55% would follow an AI tool to a better deal. Loyalty membership ranked last as a reason to trust an AI recommendation.

Non-travel companies are also outbidding on rewards. The same Expedia hotel inventory returns 1% to a free OneKey member and 10% in credits to an Uber One member. A points discount is a price, and agents compare prices. Skift Research recommends on-property recognition — the part of loyalty an algorithm cannot list.

Are Airlines Rebuilding the Shelf for Agents?

Agents need offers they can change and cancel, not just fares. Delta plans to launch NDC content (New Distribution Capability, the airline retailing standard for rich offers) by year-end, later than American and United — a marker of how unevenly airlines are moving toward content agents can service.

Where the Decisions Land

Most of these positions will harden before 2027 budgets lock. Skift Commerce Summit convenes the executives making those calls on November 11 at Apella by Alexandria in New York City, with an opening hotel-versus-OTA debate on what an OTA booking is worth to a hotel owner, and a closing audience vote on whether to build AI-ready capabilities before returns are clear.