Booking Holdings CEO Glenn Fogel disclosed during an earnings call that his company is among the first participants in the test. Google has also been working with Expedia and major hotel companies including Marriott, Wyndham and IHG as it develops its approach to agentic travel booking. The test remains limited, and Google has released relatively few details about exactly how much of the transaction is currently being handled within AI Mode. Still, it represents an important progression from using generative AI primarily to answer travel questions toward allowing software agents to participate directly in the process of shopping for and booking a hotel.
Google is building the infrastructure needed to support that progression through Universal Commerce Protocol for Lodging, or UCP. The open standard is being designed to allow hotel suppliers to turn interactions on AI surfaces such as AI Mode in Search into direct room reservations, including a real-time price and availability check before a traveler reaches the final booking step.
UCP for Lodging is still under development, and Google is currently inviting prospective partners to join a waitlist while more detailed onboarding specifications are prepared. The direction, however, is already clear: Google wants hotels and travel companies to be able to support bookings without forcing a traveler to abandon an AI conversation and restart the process on another website.
One feature of the model should attract particular attention from hotel companies. Google says the lodging supplier remains the merchant of record, retains ownership of the customer relationship and booking data, controls the post-booking experience and applies its own terms and conditions to the reservation. That structure could make agentic booking look different from the traditional OTA model. A traveler might never navigate through a hotel’s conventional website or booking engine, yet the hotel could still receive the reservation and customer information in much the same way it would with a direct booking. Google also says participation in UCP will not influence how a hotel’s property or rates rank in its listings. Offers connected through UCP can provide a direct booking option, but adopting the protocol itself does not provide a ranking advantage.
The economics will matter just as much as the technical structure. A hotel connected directly to an AI booking experience through its own systems could face a very different acquisition cost from one whose rates reach the same traveler through an OTA, GDS, representation company or other intermediary. That question becomes more complicated as established hotel distribution companies become involved. Amadeus says it is working with Google and other hospitality companies as a foundational partner to co-develop UCP for Lodging while also building technology to help hotels participate in AI-powered travel marketplaces.
Amadeus already operates reservation and distribution infrastructure used by hotels to move rates and inventory across numerous sales channels. Its involvement suggests that agentic booking may develop partly by extending existing hotel distribution technology into new AI interfaces rather than replacing the industry’s current infrastructure wholesale. That could be especially important for independent properties and smaller hotel groups. Large hotel brands have the technology resources and scale to work directly with companies such as Google, while smaller operators are more likely to rely on their CRS, booking engine, channel manager or connectivity provider to reach new distribution channels.
Google already operates a large network of hotel connectivity partners that send rates to Google for free booking links and Hotel Ads. Extending a similar connectivity model to agentic booking could eventually allow a large number of independent hotels to participate without creating their own direct integrations with every AI platform.
At the same time, Google is changing the way travelers can search for hotels before they ever reach the booking stage. Its Ask Maps feature can now interpret conversational requests involving not only destination, dates and price but also less structured criteria such as atmosphere, proximity to restaurants and gyms, and the context of a particular trip.
Google offered the example of a traveler looking for a reasonably priced, highly rated hotel with an artsy atmosphere near a conference in downtown Miami and within walking distance of a gym and restaurants. Ask Maps can evaluate that request using real hotel pricing and availability rather than simply returning a conventional list based on a handful of filters. The current Ask Maps hotel experience stops short of completing the reservation and directs travelers to partners to book. Google is nevertheless developing the pieces that could eventually connect conversational discovery, real-time hotel availability and an embedded transaction much more closely.
Personalization could deepen that change. Travelers can opt to connect Ask Maps with Gmail so the system can use information about existing flights, hotel reservations and other plans, with additional integrations such as Calendar expected later. Hotel discovery could therefore become considerably more contextual than conventional search. Instead of entering a destination and dates and then applying filters, a traveler could describe the purpose of a trip, desired neighborhood, price range, preferred atmosphere and nearby activities and allow an AI system to assemble the shortlist. That creates another challenge for hotel digital marketing teams. Properties have spent years optimizing websites and listings for search engines, but conversational discovery depends heavily on whether an AI system can accurately understand a hotel’s amenities, location, room types, policies, pricing, images, reviews and other attributes.
Structured and accurate hotel data becomes even more valuable in that environment. If an AI assistant is asked to find a quiet boutique hotel near a convention center with a good fitness facility, late-night dining nearby and suitable rooms for a family, every piece of reliable information that helps distinguish one property from another can affect which hotels make the shortlist. Amadeus is already developing products around this emerging form of discovery. Its recently announced Performance Manager – AI Search is intended to help hotels improve their visibility in AI-generated results, while the company is also developing an AI Booking Assistant for hotel-owned websites and its iHotelier booking engine.
Amadeus reported that in early pilots, 44.7% of visitors arriving through AI-driven sources reached a hotel booking engine compared with 25.9% of visitors coming from organic sources. Those are Amadeus’ own early findings rather than an industry benchmark, but they suggest that travelers who have already used an AI assistant to narrow their choices may arrive with relatively strong booking intent.
The competition around agentic hotel booking extends well beyond Google. Booking Holdings and Expedia have enormous inventories and established consumer relationships, while hotel companies such as Marriott, IHG and Wyndham have strong incentives to ensure that AI-driven travel planning does not weaken their direct booking and loyalty strategies.
Hotel technology providers also see an opportunity to become part of the infrastructure connecting properties with AI agents. Sabre, for example, has introduced agent-ready APIs and its own Model Context Protocol server intended to let AI applications shop, book and service travel, although its publicly disclosed work should not be confused with Amadeus’ role in developing Google’s lodging protocol.
The bigger change is occurring in the booking path itself. Hotel shopping has traditionally required travelers to move through a series of separate interfaces: search for a destination, compare properties, click into an OTA or hotel website, choose a room and rate, enter guest information and then complete the transaction. Agentic systems are designed to reduce that sequence. A traveler could instead describe a trip conversationally, refine the requirements, compare a small number of recommendations and eventually authorize the system to carry out more of the booking process.
Reducing those steps could improve conversion by eliminating some of the friction that causes travelers to abandon reservations. It could also make the platform conducting the conversation more influential because the traveler may see only a handful of hotels selected by the agent rather than dozens of properties presented on a traditional results page. That prospect gives hotel companies good reason to watch Google’s role carefully. Google has said it does not intend to become an OTA, and its UCP model leaves the hotel supplier as merchant of record and owner of the customer relationship, but the company could still exert enormous influence over which hotels travelers discover and ultimately consider booking.
There is also no guarantee that every route into agentic booking will have the same economics. Hotels will need to determine whether reservations are arriving directly or through another distributor, what commissions or transaction fees apply, whether loyalty benefits can be recognized and how guest information moves into the hotel’s existing systems. Those questions are familiar to hotel distribution executives even if the interface is new. Hotels have spent decades weighing the reach provided by intermediaries against the cost and customer ownership advantages of direct channels, and AI-driven booking is likely to create another version of that same tradeoff.
The technology is still at an early stage. Google’s current hotel booking test reaches only a limited share of U.S. users, UCP for Lodging is still being developed, and conversational hotel search in Ask Maps does not yet provide a complete end-to-end booking experience. What has changed is that agentic hotel booking is no longer merely a concept being discussed for some distant future. Google is testing it with major travel companies, building a commerce protocol specifically for lodging and connecting conversational hotel discovery with live rates and availability.
Hotel companies now have another distribution channel to prepare for, even if its final shape remains uncertain. The properties that understand how their inventory, content, pricing and guest data will move through AI-driven booking environments will be in a better position to decide where those bookings should come from, what they should cost and how much of the guest relationship they are willing to place in the hands of a new generation of travel intermediaries.
By Dustin Stone
