What direct AI hotel distribution looks like
A handful of hotels and groups are already building it. Here is what they have in common
Two months ago, a hotel that wanted a direct presence inside ChatGPT had no real path to get there. Booking.com and Expedia had secured their places as launch partners. The OTAs moved fast, as they always do. Hotels watched.
That has changed, visibly, in the past six weeks. The question is no longer whether direct AI distribution is possible. It is whether any given hotel is building it or waiting to see how it goes.
The ones building it fall into roughly three categories. Each is doing something meaningfully different. The differences matter.
The chain that treated AI as infrastructure
Hyatt is the clearest example of a hotel company that made a decision — not a pilot, a decision — to rebuild around conversational search before the channel existed.
The work started in early 2025 with the redesign of Hyatt.com's search experience. The premise was straightforward: travelers increasingly describe trips the way they would to a person, not the way they would fill out a form. The old city-date-availability interface is not built for "I want a quiet resort in the mountains for a long weekend in April." The new one is.
Hyatt has now published what it calls longitudinal data across multiple quarters. Booking conversion rates through the new search are measurably higher. Total revenue from the channel is up. The same AI investment that rebuilt the website also deployed tools across Hyatt's group sales function — and group sales productivity has improved by nearly 20%.
In February 2026, Hyatt announced a branded ChatGPT app. The model is consistent with the website rebuild: AI surfaces the hotel, the traveler books on Hyatt's own platform. No in-chat checkout, no intermediary taking a margin on the transaction. The AI is the front door. The direct channel is the building.
This is what it looks like when a company treats AI as a long-term infrastructure shift rather than a feature to be tested and reported on at the next board meeting.
The group that moved first on the open platform
Accor launched the ALL Accor app inside ChatGPT on January 29, 2026, becoming the first major hotel group with a native presence in the world's most-used AI platform.
The mechanics are worth noting. A traveler can search across Accor's entire portfolio — more than 5,700 hotels across 45 brands — using natural language. They can see both public rates and ALL loyalty member pricing inside the conversation. When they are ready to book, they are directed to Accor's own platform. The transaction does not complete inside ChatGPT. Accor is the merchant of record.
That last part is the strategic point. Accor is using ChatGPT as a discovery surface, not ceding the booking relationship to it. The experience reinforces the loyalty programme at the moment a guest is deciding where to stay — not after they have already clicked through to an OTA.
Accor is also doing something that most hotel companies have not figured out yet: making its member pricing visible inside the AI conversation. For a traveler comparing Accor against a competitor, seeing that a loyalty rate exists — before they have even reached the booking engine — changes the conversation.
The integration covers more than 20 languages. It is available in all markets where ChatGPT supports third-party apps. This is not a regional experiment.
The negotiated lane
Marriott took a different route to the same destination. Rather than building for an open AI platform, it negotiated directly with Google.
In February 2026, CEO Anthony Capuano confirmed on the company's Q4 earnings call that Marriott was already working with Google to develop a "priority search experience" inside AI Mode. The phrasing matters: not general eligibility, not participation in an auction, but a priority lane — a negotiated position within the AI surface that the majority of the market cannot access on the same terms.
This is the Marriott playbook applied to a new distribution environment. The chain is large enough, and its loyalty base valuable enough, to negotiate directly where others compete on price. It is a strategy available to very few hotels.
What it confirms, more than anything, is that Google's AI Mode is not neutral infrastructure. It is a surface with tiers, and the top tier is being allocated through relationships rather than auctions. Whether that changes as AI Mode scales is an open question.
The path for everyone else
The challenge with the three examples above is that none of them is a template for the independent hotel or the mid-scale regional group. Hyatt rebuilt its technology stack over multiple years. Accor has 5,700 hotels to justify the development cost of a ChatGPT app. Marriott negotiated at a scale that is not replicable.
On March 4, 2026 — last week — Lighthouse launched The Hotels Network app inside ChatGPT. It is not the first direct booking solution for independent hotels in an AI platform — Selfbook enabled hotel booking inside Perplexity in March 2025, covering roughly 140,000 properties. But the Lighthouse product is the first app inside ChatGPT's App Directory that is open to hotels of any size without a custom integration, built on the Model Context Protocol standard and available for a flat subscription fee with no commission on bookings. Given that ChatGPT is the dominant AI platform by user volume, that distinction matters.
The model is straightforward. A hotel connects through MCP. Its verified content — official descriptions, real rates, live availability — feeds directly into ChatGPT. When a traveler discovers the property through the app, they are routed to the hotel's own website to complete the reservation. The hotel is the merchant of record. Lighthouse takes the subscription, not a cut of the booking.
Hotels connecting through the app include Minor Hotels, One&Only Resorts, and Preferred Hotels — groups with the scale to build their own solutions but choosing a faster path to market. The same channel is available to the independent property with one front desk.
The practical question for any hotel not in the Marriott tier is this: if a traveler asks ChatGPT for a hotel in your destination tonight, what does the response look like? In most cases, the answer is scraped OTA listings, generic summaries, outdated information, and a link-out to Booking.com. That is the default. The hotels connecting to MCP-enabled channels are the exception.
What they have in common
The approaches differ. The underlying logic is the same.
Every hotel in this piece is doing three things: feeding verified, structured data directly to AI platforms rather than relying on scraped summaries; keeping the transaction on its own platform rather than inside the AI interface; and treating the AI discovery layer as a new front door to an existing direct relationship, not as a new intermediary to be managed.
None of them is completing bookings inside the AI. None of them is ceding merchant-of-record status. None of them is paying a commission on the outcome.
The question that has no answer yet is volume. AI travel planning is growing fast, but it is still early. No hotel has published meaningful booking numbers attributable to AI discovery channels. The infrastructure is being built before the demand has fully arrived.
That is exactly what happened with mobile. The hotels that built mobile-optimised booking engines before mobile traffic became dominant captured the channel cleanly. The ones that waited adapted on the OTAs' timeline.
The hotels worth watching are the ones that are not waiting to find out which version of this story repeats.
by Markus Busch, Editor/Publisher Hospitality.today
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 →