The old hotels-versus-OTAs fight matters less than the one they now share: being found and booked by AI. Hotels win more direct bookings when their rates, policies, and guest data are unified, owned, and current enough for an AI to trust.
Hotels have spent a decade building the tools to take bookings directly. OTAs still win the demand. The 2026 State of Distribution report, drawn from 270-plus hotel brands across 53 countries, finds that core direct-booking infrastructure is now nearly universal, yet OTAs generate close to twice the bookings of hotels’ own digital channels.
The OTA is hard to leave for good reasons. It reaches international and targeted demand a single brand cannot buy on its own. It fills rooms in off-peak and last-minute windows. On many bookings it collects the payment and absorbs the fraud and chargeback risk it would otherwise carry.
To that, add a competence now becoming decisive: the OTA runs the most disciplined inventory operation in travel, holding each room as structured data, its rates, policies, and attributes in uniform fields a machine can read directly rather than a marketing page it has to interpret. Keeping that content consistent used to count as back-office housekeeping. As AI agents begin to mediate the booking, it becomes a distribution advantage.
The shift is already under way. The same report finds AI-originated search now produces a measurable share of reservations, even as 55% of hotels report little or no change to their distribution strategy.
The intermediary OTA earns its keep
Hotels keep the OTA because it pays. A 2025 Cornell Hospitality Quarterly study of 644 U.S. hotels’ income statements tied OTA participation to higher occupancy, RevPAR, and EBITDA. The profit benefit thinned only for economy properties. Direct booking, for most of these operators, has been a way to improve margin on top of a channel that already works, not a way to replace it.
So the useful question for revenue and distribution leaders is narrow. A traveler who means to book direct often lands on the OTA anyway, because the OTA ranks first on the hotel’s own name and, increasingly, is what an AI assistant surfaces for a direct query. The contest is which channel captures that booking, and at what cost to the guest relationship.
What the OTA actually sells
Strip away the traveler-facing brand and the OTA’s real product is consistency: one description, one rate, one set of policies, the same everywhere. Travelers reward it, and machines increasingly require it. That operation is concentrated. In Europe, where distribution data is best documented, HOTREC finds Booking Holdings and Expedia Group together control more than 85% of OTA bookings, and the two dominate online hotel distribution in most major markets. A growing share of a hotel’s own legibility now sits inside two companies.
Borrowed legibility
A hotel’s own offer rarely reads as cleanly as the OTA’s version of the same room, and the reasons are varied. A room’s attributes sit scattered across property-management, reservation, and content systems that never reconcile. Cancellation and rate rules stay buried in rate-plan and loyalty logic rather than exposed as structured terms. The room gets merchandised as marketing prose rather than the facts a traveler or an agent looks for. More than half of European hotels report undercut rates as per HOTREC, sometimes from the brand’s own promotions, sometimes from rogue intermediaries reselling wholesale inventory through API feeds the hotel never meant to open.
Underneath it sits a data problem, and not only for small operators: the same State of Distribution research finds hotel commercial teams still working across disconnected systems and disparate data. Without one live view of rates, inventory, and policy, prices drift out of line, margin leaks through the inconsistencies, and the offer a hotel publishes stops matching the stay it can deliver. That is the inconsistency an agent is built to catch.
What reads as annoyance to a traveler reads as a signal to an agent, which reaches the cleanest, most complete offer it can find. Today that is frequently the OTA’s. The plumbing is already in place: OTA apps from Expedia and Booking.com now run inside ChatGPT with live access to their own rates and inventory, while most hotels’ direct inventory is not. When an AI assistant surfaces the room, the guest books through the OTA’s app, and the reservation and commission land there, exactly as if the guest had started on the OTA. The hotel pays to win a guest it could have booked directly.
Brand equity helps only when it is machine-readable. Phocuswright’s read is that hotels with strong brands, loyal bases, and proprietary content are best placed as AI moves into booking: brand equity routes the booking when it shows up as machine-readable facts and a stay the hotel can reliably deliver. Recognition on its own stays invisible to the agent.
What the agent weighs beyond price
Rate parity between hotels and OTAs is still contested; the undercutting numbers show the fight is far from settled. But a second front is opening, over the terms around the rate. A 2025 study in Tourism Economics found that when prices match, travelers lean toward the direct site; when prices diverge, cancellation flexibility and channel credibility start to decide the booking, both of which favor the OTA. Free cancellation with no upfront payment reads almost like a free option, which travelers value immensely.
An agent applies these judgments at scale. It reads cancellation terms and descriptions directly, so an offer it cannot interpret gets passed over. Whether a property honors what it advertised, the agent infers as a careful traveler would: from reviews, ratings, and the record of bookings that completed cleanly. An overbooking or a billing error shows up to a system reading that trail. The promise has to match the stay, because the agent scores both.
Learn the intermediary’s discipline
The move for hotels is to build the OTA’s discipline themselves, and that work runs deeper than the storefront. It starts with a single source of truth for inventory, rates, policy, and guest history. That data has to be live, open to every commercial team, and auditable, so the same facts drive the website, the agent-facing offer, pricing, and OTA reconciliation. Structured cancellation terms and room facts on the page, rather than prose or a PDF, are the visible half. The larger half is the backend that feeds them: clean, connected data is what lets a hotel price accurately, run efficient operations, recognize a returning guest, and get full value from the OTA relationship rather than just surviving it.
The organizational part is harder. Guest and operational data sits across brand, owner, operator, and property, and the brand has to own one legible, deliverable offer across all four, because the brand is what the guest holds responsible when the stay goes wrong. The gap runs wider than hospitality. Deloitte’s State of AI in the Enterprise 2026 finds that legacy data architectures cannot power autonomous AI, and that a trusted data strategy built on domain-owned data products is the precondition for agents to deliver. Agent-ready visibility is one return on that foundation. Sharper pricing, smoother operations, and durable guest loyalty are the others.
Own the offer a machine can trust
The direct-booking fight was always about who the hotel pays to reach its own guest. The agent layer narrows it to one question: whose version of the room reads cleaner to a machine. The hotels that own their data layer, and the legible, deliverable offer it produces, are the ones positioned to hold the guest relationship and the margin as agents take on more of the booking.
About This Analysis
This Incisiv analysis draws on current industry benchmarks. Global channel, data-fragmentation, and AI-search figures come from The State of Distribution 2026 (RateGain, NYU SPS Jonathan M. Tisch Center of Hospitality, and HEDNA; more than 270 hotel brands and 58,000 properties across 53 countries). European distribution figures, including OTA concentration and undercutting, come from HOTREC’s European Hotel Distribution Study 2026. The AI distribution framing draws on Phocuswright’s U.S. hotel and lodging 2026 research, with current AI-booking integrations reported by Hotel Dive. Profitability findings are from a 2025 Cornell Hospitality Quarterly study of 644 U.S. hotels’ income statements; non-price booking behavior from a 2025 Tourism Economics study. Enterprise data-readiness context is from Deloitte’s State of AI in the Enterprise 2026.
