Universal Cart Just Added Hotels and Food Delivery. UCP Isn't Just Retail.

Google quietly stretched Universal Cart past retail this week. Booking and Expedia are going into AI Mode for hotel reservations. Food delivery is heading into Google Maps conversations. The same protocol that powers Nike, Sephora, and Target now covers nights, meals, and rides. If you assumed your category was further out from agentic commerce, May 20 reset the clock.

TL;DR: At Google Marketing Live 2026, Google extended the Universal Commerce Protocol (UCP) beyond retail into hotels (Booking, Expedia) and food delivery. Google also announced Conversational Attributes, a new product-data field type, and AI Performance Insights for share-of-voice tracking. The bottleneck for every category, retail or services, is the same: structured catalog data that AI agents can read and rank.

What Actually Changed at Google Marketing Live 2026?

Google Marketing Live 2026, held May 20, was the merchant-tooling layer underneath the Universal Commerce Protocol (UCP). Six concrete announcements landed in one keynote.

Booking.com and Expedia confirmed integrations into AI Mode for hotel discovery and reservations, with food delivery extending the same protocol into Google Maps conversations (Search Engine Land, May 20). Google also launched Conversational Attributes, a new product-data field set inside Merchant Center that brands fill in to match how shoppers phrase questions to AI (Search Engine Land, May 20). Affirm and Klarna are now embedded directly inside Google Pay for Universal Cart, closing the consumer-side friction in the agentic checkout flow (Google blog, May 20).

The vertical expansion is the biggest move. UCP was framed in March 2026 as a retail protocol co-built with Shopify. Two months later it covers Marriott-tier hotel bookings and food delivery.

Key stat: AI Mode now has 1 billion monthly active users about a year after launch, and AI Mode searches run 3x longer on average than traditional Search queries. Brainstorming-style queries like "ideas for a long weekend in Mexico City" are growing 30% faster than AI Mode overall (ppc.land GML recap, May 20).

Why Hotels and Food Delivery Sit on the Same Protocol as Nike

UCP was built around three primitives: catalog, cart, and identity linking. None of those primitives are retail-specific. A hotel room has SKUs (room types, dates, occupancy). A food delivery order has SKUs (items, modifiers, pickup windows). Identity linking matters for loyalty in every category. Once the protocol works for one product type, the marginal cost of adding another is mostly mapping fields.

That's the technical answer. The strategic answer is different. Google needs UCP to be the default surface where AI shopping happens before OpenAI's Agentic Commerce Protocol (ACP) can entrench inside ChatGPT. Pulling Booking and Expedia in week ten gives UCP travel queries, which carry some of the highest intent and margin on the web. Food delivery into Google Maps gives Google a local-services foothold against DoorDash and Uber Eats.

The takeaway for any brand not in the retail launch group: your industry's protocol on-ramp may be much closer than the March announcement implied. If hotels and food delivery clear the integration bar, beauty subscriptions, healthcare retail, auto parts, and B2B distribution are all addressable inside the same standard.

What "Conversational Attributes" Actually Asks Brands to Do

Conversational Attributes are new Merchant Center fields that mirror how shoppers ask AI questions, not how they query a search box. "Travel-friendly hair dryer for a 3-week Europe trip" is not a keyword string. It is a sentence with context, constraints, and intent.

This is the official Google version of what AI surfaces have been quietly demanding for a year: richer, more context-aware structured product data that goes beyond title, price, image, and category.

The product-card gap most brands see in AI visibility monitoring is largely a Conversational Attributes problem in disguise. A product that gets mentioned 100% of the time but only shows as a card 48% of the time is usually missing the contextual fields that let AI confidently render it with image, price, and stock. Google just gave the gap an official name.

How AI Performance Insights Reframes the Measurement Question

AI Performance Insights is a new Merchant Center reporting tool that benchmarks a brand's share of voice against similar competitors across AI Mode, Gemini, and AI Overviews (Search Engine Land, May 20). Rolling out in the US, Canada, Australia, India, and New Zealand "in the coming months."

This is Google productizing AI share of voice measurement natively, but only inside the Google ecosystem. It tells you how you rank on AI Mode, Gemini, and AI Overviews. It does not tell you how you rank on ChatGPT, Perplexity, Microsoft Copilot, or Amazon's Alexa for Shopping. For a hotel brand whose customers ask travel questions across every assistant, "Google-only" is a partial view. Tools like Paz's AI Readiness Report close that gap by measuring across multi-channel AI commerce surfaces.

Either way, "how are we doing on AI?" is no longer a vague boardroom question. It now has a number. Brands that wait for the dashboard to turn red will be six months behind brands that fixed their data first.

What the Vertical Expansion Means for Mid-Market Retailers

Three implications worth holding onto.

First, the discovery layer is collapsing across categories. A shopper planning a Mexico City trip can ask AI Mode about flights, hotels, restaurants, beach gear, and food delivery in one conversation. Brands in any of those categories sit in the same retrieval pool. The data quality required to surface in a hotel query is the same as for a sunscreen query: dense conversational commerce attributes, image-anchored cards, and identity-linked offers.

Second, the timeline compression is real. Vogue framed it bluntly: "Brands auditing those attributes now are quietly setting themselves up to be the ones Universal Cart recommends. The ones ignoring them are about to find out how expensive that decision was" (Vogue, May 20). Forbes ran a parallel piece the same day calling Universal Cart the official start of "a new era of e-commerce in which AI shapes how online shoppers discover, engage with, and purchase products" (Forbes, May 20). When tier-one consumer business and luxury fashion outlets both treat AI catalog readiness as a brand-side urgency, the C-suite conversation has shifted.

Third, "wait and see" is now visibly expensive. Delaying AI catalog work by a quarter also means skipping Google's free AI Performance Insights baseline, the Conversational Attributes early-mover window, and the BNPL-native Universal Cart placement.

What to Do This Week

1. Pull your Merchant Center feed audit. Check whether you can populate the new Conversational Attributes fields with non-placeholder values today. Where you can't, identify the catalog or PIM gap and assign it an owner. Reference the official Google Merchant Center for AI Mode guide.
2. Run an AI visibility baseline across more than Google. AI Performance Insights, when it ships, will only tell you what Google sees. Get an external agentic storefront and cross-platform read on ChatGPT, Perplexity, and Microsoft Copilot before the Google-only number becomes the only number anyone looks at.
3. Decide which UCP TAM you're in. If you sell hotel nights, restaurant orders, services, subscriptions, or B2B parts, you are no longer outside UCP's scope. Map the protocol fields to your category and identify which ones are missing in your current feed.
4. Stress-test the AI search queries that matter for your category. A travel brand should be testing "best hotels for X" and "things to do in Y" the way a beauty brand tests "best foundation for combination skin." If your products don't surface as product cards, the catalog management layer is the gap.
5. Brief leadership on the Vogue and Forbes framing. Mid-market brand presidents and CMOs read those outlets. The window where AI catalog readiness is a "tech project" and not a "board topic" has closed.

Frequently Asked Questions

Is Universal Cart only for retail brands?

No. As of Google Marketing Live 2026 on May 20, Universal Cart and UCP extend to hotel reservations through Booking and Expedia inside AI Mode, and to food delivery through Google Maps conversations. The protocol primitives (catalog, cart, identity linking) work for any category with discrete units of inventory, not just retail SKUs.

What are Conversational Attributes in Merchant Center?

Conversational Attributes are a new product-data field type Google announced May 20, 2026. Brands populate them so AI surfaces can map natural-language shopper questions to specific SKUs. They sit alongside existing Merchant Center attributes and are explicitly designed for AI Mode, Gemini, and AI Overviews discovery, not classic keyword search.

Does Google's AI Performance Insights replace third-party AI visibility tools?

Only inside the Google ecosystem. AI Performance Insights covers AI Mode, Gemini, and AI Overviews share of voice. It does not measure ChatGPT, Perplexity, Microsoft Copilot, or Amazon's Alexa for Shopping. For a full agentic visibility picture, brands need a cross-platform read alongside the Google-native baseline.

How fast should a non-retail brand start preparing for UCP?

The hotel and food-delivery integrations announced May 20 went from "future" to "rolling out" inside the same protocol that launched in March 2026. That is a 60-day cycle. Any brand with a structured catalog should be auditing feed quality and identity-linking readiness in the current quarter, not next year.

Does Affirm and Klarna being inside Google Pay change anything for merchants?

Yes. Previously BNPL inside Universal Cart required a redirect step. With Affirm and Klarna embedded directly into Google Pay, the friction inside the agentic checkout drops, and merchants enrolled in Universal Cart inherit BNPL natively. The merchant still owns the post-purchase relationship, since the transaction completes on the retailer's checkout, not inside the AI surface.

The story from May 20 is not the headline launches. It is that UCP graduated from a retail standard to a category-agnostic protocol in two months, with Conversational Attributes and AI Performance Insights as the matching tooling. The data work brands have been delaying for a year is what now decides who shows up when shoppers ask AI to plan their trip, fill their cart, or pick their dinner.