Google Marketing Live 2026 Extends UCP to Hotels and Food — Agentic Commerce Moves from Buying Products to Booking Services
Google Marketing Live 2026 extended the Universal Commerce Protocol to hotel booking and food delivery, with Amadeus, Booking.com, Hilton, Marriott, DoorDash, Uber Eats, Toast and Square as vertical co-developers. We read the announcement as agentic commerce's move from products to services: multi-dimensional inventory data, merchant-of-record design, and the operator-side agent Ask Advisor.
Key Takeaways
- At Google Marketing Live 2026 on May 20, Google extended the Universal Commerce Protocol (UCP) beyond retail, announcing vertical specifications for hotel booking and food delivery. Amadeus, Booking.com, Expedia Group, Hilton, Marriott and Trip.com are Lodging co-developers; DoorDash, Uber Eats, Toast and Square are Food co-developers
- Agentic commerce, which so far has handled product purchases, is stepping into service bookings whose price changes with dates and rate plans. Google's design turns AI Mode in Search and Google Maps conversations into order channels while the business remains the Merchant of Record
- Whether merchants can expose complex commerce data, such as inventory, pricing and options, in machine-readable form becomes the precondition for receiving agent-driven orders. With the operator agent Ask Advisor announced the same day, buy-side protocol and sell-side operations are converging on the same foundation
The shopping protocol steps outside retail

New UCP-powered features and AI tools will create a more intuitive, agentic shopping experience for consumers and retailers.
blog.googleAt Google Marketing Live 2026 on May 20, 2026, Google announced that the Universal Commerce Protocol (UCP) is expanding into hotel booking and local food delivery. The official blog post, published under VP/GM of ads and commerce Vidhya Srinivasan, says consumers will soon be able to book their next hotel right from AI Mode in Search, or order food delivery from a conversation in Google Maps. If Universal Cart, unveiled at Google I/O the day before, was the consumer-experience reveal, Marketing Live was the declaration that the protocol's scope now crosses industries.
UCP is an open standard for exchanging everything from catalog queries to checkout between AI agents and businesses. We cover the protocol's design and its retail participants in a separate explainer, so this piece focuses on one question: what does it mean that the protocol has moved beyond retail? The official portal ucp.dev now lists Lodging and Food alongside Shopping, with detailed specifications for both announced as coming soon.
UCP for Lodging tests multi-dimensional commerce data
One pillar of the expansion is UCP for Lodging, the hotel booking vertical. ucp.dev describes its purpose as follows.
Enables high-quality booking flows within AI surfaces — complete with real-time pricing and availability checks, smooth handling of complex rate plans, easy guest registration, and secure checkout.Source: ucp.dev / UCP for Lodging
Why does hotel booking need its own specification? Product inventory can largely be expressed as a binary: in stock or not. Hotel inventory changes price with dates, room types, occupancy, cancellation policy and breakfast inclusion. The same room becomes a different product with every combination of conditions. That is multi-dimensional commerce data. For an agent to handle a booking, this structure must be machine-readable in real time, and a retail product catalog spec alone cannot carry it. The vertical specification is positioned as the extension that absorbs that gap.
The lineup of co-developers is just as notable. The Lodging section of ucp.dev lists Amadeus, the largest GDS (the distribution systems behind airline and hotel inventory), the OTAs (online travel agencies) Booking.com, Expedia Group and Trip.com, the direct-booking-focused chains Hilton and Marriott, and Google. Companies that have long fought over inventory and customers, and the GDS layer that connects them, are now sitting at the same specification table. Skift reported that Google has publicly committed to extending its agentic commerce infrastructure to travel.
The entry point for the experience is AI Mode in Search. Google's lodging developer page says the protocol will enable direct, instant reservations from an AI Mode conversation. UCP is designed to interoperate with existing standards, including AP2 (Agent Payments Protocol, which governs an agent's payment authority and limits), A2A and MCP, and is built on industry standards such as OAuth 2.0. The detailed booking-flow specification and onboarding steps have not been published yet; at this stage Google is accepting waitlist registrations from prospective participants.
The design decision to leave Merchant of Record with the business
One theme ran through the entire announcement: who owns the transaction. The lodging developer page states plainly that "You remain the Merchant of Record. Keep all of your customer data and relationships," and that UCP creates a transparent accountability trail between businesses, credential providers and payment services (Google). The official blog likewise stresses that regardless of how a shopper chooses to buy, the retailer always remains the merchant of record.
The Merchant of Record is the legal seller of a transaction, responsible for collecting funds, issuing refunds and meeting compliance obligations. Losing that position in agent-mediated sales would mean surrendering the customer relationship and its data to an intermediary. This is the fork in the road: does the pattern seen with marketplace dependence in e-commerce and OTA dependence in travel, commissions paid and customer contact lost, repeat itself at the AI layer? On paper at least, UCP chose to leave the transaction and the customer data with the business. We map this broader debate in our piece on open versus walled-garden agentic commerce.
Even so, keeping Merchant of Record status does not change the fact that discovery moves onto Google's surfaces. Which properties or stores the agent surfaces as candidates, how exposure is determined, and what commercial terms play the role commissions play today, all await the detailed specifications and contractual terms. It is safest to evaluate the published design philosophy and the still-unsettled operating conditions separately.
UCP for Food makes customization machine-readable
The other vertical, UCP for Food, is described on ucp.dev as follows.
Powers conversational food ordering journeys that seamlessly handle nuanced meal customization, real-time availability and deals, tipping, and delivery instructions through a scalable checkout experience on AI surfaces.Source: ucp.dev / UCP for Food
If lodging's hard problem is multi-dimensional rates, food's is combinatorial customization and immediacy. Settling toppings, spice levels, delivery address and time, and tipping inside a conversation requires the store's menu and options to be structured data. The co-developers include DoorDash and Uber Eats, joined by the POS (point-of-sale) providers Toast and Square. That selection signals the scope covers not only the route through delivery platforms but also a direct path from store systems to agents.
The design principles mirror Lodging. The Food developer page also commits to the business remaining Merchant of Record and to tokenized payments that work with existing payment integrations. As with Lodging, detailed specs and onboarding are marked as coming soon.
Sell-side operations also go agent-first
The same day brought a major sell-side announcement. Ask Advisor, embedded in Merchant Center, is an agent that surfaces insights across Google Ads and Google Analytics and executes tasks such as campaign setup. The beta is available for English-language accounts (Search Engine Land).
On the measurement side, Google introduced AI Performance Insights, which visualizes how often your brand appears on AI surfaces through a share-of-voice comparison against similar brands, rolling out in the US, Canada, Australia, New Zealand and India. Conversational Attributes, announced alongside it, lets merchants adapt product descriptions to the more conversational way people now search (Google).
The structure matters more than any individual tool. The buy-side protocol (UCP) and the sell-side operations agent (Ask Advisor) are converging on the same Merchant Center data foundation. In an era when agents do the buying, agents also handle the seller's analytics and ad operations. That symmetry is the thread running through Marketing Live as a whole.
What retail and e-commerce operators should take from this
At first glance, hotel and food verticals sit outside product e-commerce. The capabilities being tested, however, are the same.
First, the scope of commerce data that must be machine-readable is widening. Even in product retail, whether an agent can receive service-like attributes, delivery time windows, gift options, subscriptions, warranty and return conditions, will increasingly decide whether an order can happen at all. The Lodging spec, built for multi-dimensional rates, is a preview of how far the machine-readability bar will rise for complex commerce data. We cover how to approach that work in agent-ready product data.
Channel strategy also needs a fresh look. If AI Mode and Google Maps become order channels, operators must design how traffic and measurement are allocated across their own site, marketplaces and intermediaries, and agent-mediated flows. Retaining Merchant of Record status and customer data softens the dependence risk, but it does not guarantee control over exposure.
At the same time, rushing investment decisions would be a mistake. Detailed specifications for both Lodging and Food are unpublished, and participation is at the waitlist stage. The realistic sequence is to separate confirmed specifications from announced design philosophy, then estimate integration costs against your own systems once the specs are out.
Conclusion
The real significance of Google Marketing Live 2026 is that it marked agentic commerce's turn from buying products to booking services. Hotel inventory that changes shape with dates and conditions, and food orders whose combinations keep multiplying, now get vertical specifications to make them machine-readable, presented together with a design that leaves Merchant of Record status with the business. In parallel, Ask Advisor pushes the seller's operations toward agents, sketching a structure where both sides of the market run on the same foundation.
Three things are worth watching next: when the detailed Lodging and Food specifications are published and what they contain, where the first AI Mode-driven booking or order cases emerge, and when PMS (hotel property management systems) and POS vendors put UCP support on their roadmaps. For retail and e-commerce operators, the most reliable preparation available today is to test whether your own commerce data can meet this level of machine-readability.



