Universal Cart & UCP: why AI shopping is becoming an infrastructure question for EU commerce leaders

What happens to your brand when the customer journey starts happening somewhere other than your website?

At Google I/O 2026, that question stopped being theoretical. Google introduced Universal Cart, a Gemini-powered shopping cart that follows the customer across Search, YouTube, Gmail and the Gemini app. It sits on top of two pieces of infrastructure Google has been building in plain sight: the Universal Commerce Protocol (UCP), which lets AI surfaces and merchants talk to each other, and the Agent Payments Protocol (AP2), which makes agent-led payments verifiable. Together, these are the foundations of the agentic commerce flow Google is constructing now. Google announcement

For most EU commerce leaders, this lands in one of two ways. Either it looks like a technical project for someone else on the team, or it looks like a US story that has not arrived in Europe yet. Both reads underestimate what is happening.

Universal Cart, UCP and AP2 are not just new shopping features. They are the early shape of a different commercial model, one where the customer’s path to purchase increasingly runs through an AI layer the brand does not own. Even before any of this reaches the EU at scale, the assumptions underneath it are already changing what brand, marketing and revenue leaders need to plan for.

This is what is changing, and why it matters now.

From click to buy to intent to order

The traditional e-commerce journey was built around a sequence of customer actions: search, click, website visit, product page, cart, checkout and payment. The emerging AI-assisted journey starts earlier, with intent. A customer may ask an AI surface to compare options, monitor prices, remember preferences, check compatibility or move items into a cart before reaching the brand’s own environment.

That difference may look small. It is not. In the old model, the brand had more control once the shopper arrived on the website. In the new model, more of the decision-making context may happen before the customer reaches the brand’s own environment.

What this changes for marketing across the buyer journey

For a marketing leader, the practical question is not whether the customer journey is changing. It is what changes at each stage, and what needs to happen differently in response.

Awareness. Traditional brand awareness work was about reaching the customer through paid media, organic search, social and content channels. In an AI-assisted journey, awareness increasingly happens inside the AI layer. The customer asks a question. The AI surfaces brands. The brands that get surfaced are the ones the AI system can confidently understand, classify and recommend. Awareness is no longer only about reach. It is about whether the brand is legible to machines.

Consideration. In the traditional journey, consideration was where the brand could put substantial effort into content, product detail, comparison guides and reviews. In an AI-assisted journey, the AI does much of the comparison itself. It summarises competitors, surfaces alternatives, weighs trade-offs and recommends. The brand's content still matters, but its job changes: from convincing the customer directly to giving the AI system the right material to represent the brand accurately when the customer is not even on the brand's website.

Conversion. In a traditional journey, conversion lived inside the brand's own checkout. The new conversion lever is no longer only the checkout page. It is the eligibility, trust and accuracy of the product data the AI system is operating on when it constructs the cart and prepares the transaction. A brand can lose conversion not because the website is slow, but because the product feed is incomplete or the checkout endpoint is not ready.

Retention. Loyalty and repeat purchase have traditionally lived in the relationship between brand and customer through email, app, account and direct channels. AI-mediated commerce introduces a third party into that relationship. The customer may interact with the AI more often than with the brand. The retention question becomes whether the brand's signals, preferences and history are being carried forward by the AI in a way that respects the brand-customer relationship, or whether the AI substitutes its own logic for the brand's loyalty work.

Universal Cart makes the shift visible

Universal Cart gives visible shape to something that has felt abstract for most of this conversation. It is not really a cart in the traditional sense. It is a persistent shopping layer that travels with the customer across Google’s services. Google says Universal Cart works across merchants and services, and that when users are ready to buy, UCP can make checkout smoother through Google Pay or transfer items to the merchant’s site to complete the purchase. Google also states that, no matter which way the user buys, the brand remains the merchant of record. Google announcement

That is the strategic tension Universal Cart makes visible. The brand may still own the transaction, fulfil the order and remain Merchant of Record. But the orchestration layer that connects discovery to checkout is increasingly owned by Google.

What UCP actually does

UCP is easy to misread at first. It is not a new website feature, it is not a checkout button, and it is not a switch that makes a brand agentic commerce ready. It is better understood as a commerce protocol: a shared set of rules that lets different systems exchange the information needed to support shopping and checkout.

In this context, UCP creates a more standardised way for AI surfaces, merchants, payment systems and commerce platforms to exchange the information needed to support shopping and checkout. Google says UCP gives merchants different integration paths, including native and embedded checkout, and is designed to align with existing merchant business logic. Google UCP overview

Commercially, UCP helps answer practical questions:

• Can this product be bought through an AI surface?
• Is this product eligible for native checkout?
• Is the right variant available?
• What is the correct price?
• Which taxes, shipping options and fulfilment rules apply?
• Which legal notices, warnings or policies must be shown?
• Who owns the customer relationship?
• How is payment authorised?
• How does the order flow back to the merchant?

That is why senior leaders should care. UCP is not only about checkout. It is about whether the brand’s commerce operation can be interpreted, trusted and acted upon by AI systems.

What UCP does technically, in business language

The technical documentation can look intimidating. The commercial meaning is simpler.

The human is still part of sensitive checkout

In Google’s documented native checkout flow, the user and, optionally, an agent can add items to a checkout session. When the user decides to check out, control passes to a Google UI, where the user enters sensitive fulfilment and payment details and submits the order. Google states that the agent is not involved in this manual checkout step to ensure determinism. Native checkout guide

So the immediate change is not that AI agents are completing every purchase independently. It is that AI systems are starting to shorten the distance between product discovery, comparison, cart-building and checkout, and that distance is where much of the brand-customer relationship has traditionally lived. The final confirmation may remain human, but the context around that decision may increasingly be shaped by the AI layer.

Merchant of Record does not mean business as usual

One of the most reassuring parts of UCP is that merchants remain commercially central. Google says merchants remain Merchant of Record and retain ownership of customer relationships, data and post-purchase experience. Google UCP overview

Merchant of Record status means the brand remains the legal seller responsible for the transaction. In practical terms, that usually includes the order, fulfilment, customer relationship, post-purchase experience, returns, support and transaction records. So this is not simply a marketplace model where the brand becomes an invisible supplier.

But a brand can remain the legal seller and still lose influence over parts of the commercial journey. If customers discover, compare, save, monitor and initiate checkout through an AI layer, that layer gains influence over which products are considered, how they are described, which alternatives appear and when the customer is nudged to act.

The brand may still own the transaction. But the route to that transaction may increasingly be shaped outside the brand’s own website. For commerce leaders, that is the real question of the next 18 months: not who owns the sale, but who owns the journey to it.

Why product data becomes a leadership issue

In traditional e-commerce, product data quality was often treated as an operational issue. In AI-mediated commerce, it becomes a commercial eligibility issue.

Google’s Merchant Center documentation for UCP gives a practical example. Merchants can use a native_commerce attribute to indicate checkout eligibility, and Google lists product categories that are ineligible for checkout, including certain recurring billing models, customised goods, age-restricted items, prohibited content, some digital goods and other restricted categories. Merchant Center setup

A product may not participate in certain AI-assisted checkout experiences unless the underlying data, policies and integration conditions are correct. The same applies to identifiers, variants, availability, pricing, fulfilment, returns and mandatory notices. A missing product attribute, a weak description or a mismatch between the feed and checkout API can become a blocker to visibility, selection, trust and transaction.

AP2: the trust layer behind agentic payments

UCP handles commerce interaction. AP2, the Agent Payments Protocol, addresses a different but related problem: how agent-led payments can be authorised, authenticated and audited.

Google describes AP2 as an open, shared protocol for secure transactions between agents and merchants. It is designed to address authorisation, authenticity and accountability when AI agents initiate or support payments. Google Cloud AP2 announcement

The key concept is the mandate. In simple language, a mandate is a signed digital instruction. It records what the user allowed the agent to do, under what conditions and with what limits. Google says AP2 uses tamper-proof, cryptographically signed mandates as verifiable proof of a user’s instructions. Google Cloud AP2 announcement

Agentic commerce will not scale on convenience alone. Customers need to know an agent will not spend money outside their instructions. Merchants need to know the request reflects genuine customer intent. Payment providers need accountability. Regulators will care about transparency, consent, security and responsibility.

What Google I/O 2026 changed

Before Google I/O 2026, UCP could still be described as an emerging protocol with early commerce pilots. After I/O, the picture is more concrete. UCP-powered checkout is expanding to Canada and Australia in the coming months, and later to the UK. It is also coming to YouTube in the U.S. and to additional verticals, including hotel booking and local food delivery. Universal Cart is the most visible part of this rollout, but the broader signal is that UCP is moving from pilot to production. Google announcement

For EU brands, this is not yet a signal that full UCP-powered checkout is available everywhere in Europe. But it is a signal that the direction of travel is serious. Google is connecting AI search, shopping data, cart persistence, checkout, payments, Merchant Center and advertising into a more integrated commerce environment.

Two European retailers have already moved

The most common assumption is that UCP adoption is a US story that has not reached Europe yet. That assumption is no longer accurate. Two major European retailers, in two different sectors, have publicly committed.

Carrefour. In its Carrefour 2030 Strategic Plan, published 18 February 2026, Carrefour confirmed an "unprecedented partnership with Google on agentic commerce" as part of a wider AI, Tech and Data strategy backed by €100 million in annual AI investment. Carrefour became the first food retailer in Europe to join Google's Universal Commerce Protocol. The integration is positioned around discovery and purchase through Gemini, allowing customers to find and buy Carrefour products inside Google's Gemini app rather than only through Carrefour's own e-commerce site. Carrefour had already launched Hopla+, its own Gemini-powered shopping agent, in June 2025. UCP is the next step in a deliberate AI commerce strategy, not an isolated experiment. Carrefour 2030 Strategic Plan

Zalando. Zalando has publicly endorsed Google's UCP and is positioning it as central to its AI commerce strategy. In its 2025 full-year results communication, Zalando described itself as "one of only two European launch partners for Google's Universal Commerce Protocol", building the foundation that allows customers to discover and purchase fashion and lifestyle products directly through AI chatbots like Gemini. Zalando also stated that it is already the number one referred fashion platform by conversational AIs. The UCP-powered native checkout feature in AI Mode and the Gemini app is currently limited to eligible US retailers. For European shoppers, Zalando's UCP checkout is not yet live, but the strategic commitment and the underlying integration work are in place. Zalando 2025 results

Neither company has yet published commercial results. We do not know how UCP adoption is performing for either retailer in terms of revenue, customer acquisition or category share. But the operational signal is clear. Two European competitors at significant scale, in two different sectors, have committed. The strategic conversation has shifted from "should we look at this" to "what do we know about how our competitors are positioning".

The Merchant Center foundation

Once the question becomes "how do we get ready", many marketing and commerce leaders run into the same confusion. Should they treat AI visibility as a separate strategic project, or is it the same thing as Merchant Center optimisation?

For Google's AI surfaces specifically, the two are connected. Google's documentation describes Merchant Center as the structured data layer that powers product discovery across AI Mode, AI Overviews in Search, and the Gemini app. The data feeding those surfaces is the same data feeding traditional Shopping. The work merchants have done over the past decade on Merchant Center feed quality has quietly become AI visibility work as well.

In May 2026, Google announced AI Performance Insights inside Merchant Center, a reporting suite that tracks how products are discovered across AI Mode, AI Overviews and Gemini. The four modules are share of voice, shopping funnel performance, product term insights and product attribute insights. The rollout starts in the US, Canada, Australia, India and New Zealand. Google Merchant Center documentation

Google has also been adding new feed attributes designed specifically for conversational commerce since January 2026, with several attributes carrying an explicit note that they are primarily intended for AI surfaces. The direction is consistent: the structured data layer is becoming both a search asset and an AI commerce asset, and Google is making the connection between feed quality and AI visibility explicit.

For EU commerce leaders, this changes how to sequence the work. The right framing is not "do we invest in AI visibility, or in Merchant Center". For Google's AI surfaces, it is the same investment. Strong Merchant Center data is the foundation that makes AI visibility possible. UCP then makes the products purchasable in agentic flows where the AI moves from recommendation to checkout.

The visibility question inside Google’s AI commerce stack

Visibility used to mean appearing in search results, product listings, ads or marketplaces. Inside Google’s AI commerce stack, visibility now means something more specific: whether Google can read, classify, compare and act on your product data across AI Mode, AI Overviews, Gemini and UCP-enabled checkout flows.

This is why Merchant Center matters. A product can be present in Google Shopping and still not be ready for AI-mediated commerce. It may lack the attributes needed for conversational discovery. It may have weak product descriptions, inconsistent identifiers, unclear variants, incomplete policies or missing eligibility signals.

That creates a new kind of visibility gap. The brand may appear, but not be selected. It may be selected, but not be eligible for checkout. It may be eligible, but not be trusted enough to move safely through the journey.

What remains uncertain

The direction is clearer after Google I/O 2026, but several questions remain unresolved. The current Universal Cart rollout starts in the U.S., with YouTube and Gmail to follow. UCP-powered checkout expansion has been announced for Canada, Australia and later the UK, but this is not the same as a full EU rollout. Google announcement

Consumer behaviour is also still developing. Some shoppers will welcome AI assistance for comparison, reminders and savings. Others will be cautious about delegating purchase decisions, especially where money, personal data, quality, safety or brand trust are involved. Attribution is another open question. If more of the journey happens inside AI-mediated environments, brands will need better ways to understand which content, data, offers and trust signals influenced the final decision.

Regulation will also become more important. AI-mediated commerce raises questions around transparency, consent, responsibility, recommendation bias, data sharing, platform power and liability when something goes wrong. UCP may support more structured and accountable commerce flows, but it does not automatically solve GDPR, Digital Markets Act or EU AI Act obligations.

There is also a question commerce leaders are asking everywhere but nobody has answered cleanly yet: how do brands preserve meaning, trust and differentiation inside Universal Cart and similar AI-mediated environments? When products are surfaced through an AI layer that summarises, compares and substitutes, brand value risks being flattened into specifications and price. This is one of the most important strategic questions raised by the shift to agentic commerce, and one worth preparing for separately.

Is this relevant to your brand right now?

Not every brand needs to act on agentic commerce in the next 90 days. The honest answer depends on three questions about the brand's specific situation.

1. Does your category appear in Gemini shopping queries today? Open the Gemini app. Run three or four customer-style queries in your category. If Gemini is already responding with product recommendations and competitor names, your category is being shopped through Google’s AI surfaces right now, whether you participate or not. If those queries return mostly editorial content rather than product recommendations, Google AI commerce is still upstream of your category, and you have more time.

In many premium consumer categories, AI-assisted product discovery is already visible. The commercial urgency depends on category behaviour, competitor movement and product data readiness.

2. Are any of your direct competitors moving? Carrefour and Zalando are the loudest signals. But UCP adoption is broader than those two names. If a direct competitor in your category has signed on, made commerce announcements at Google events, or is publicly working with Google on agentic commerce, the timing question is no longer abstract. If no competitor has moved, the timing window is wider, but the first-mover advantage is also still available.

3. Is your product data ready for the question, regardless of UCP? Even brands that decide UCP is not a near-term priority benefit from the underlying work. Strong product data, clear claims, consistent attributes across channels and machine-readable policies improve traditional search performance, AI search visibility, customer experience and operational efficiency. If product data is currently a known weakness, that work is worth doing regardless of UCP timing.

The pattern from these three questions:

• If your category is appearing in Gemini shopping results and a direct competitor has moved, you should be evaluating Google UCP readiness now.
• If your category is appearing in Gemini but no competitor has moved, you have a strategic choice about whether to lead or follow.
• If your category is not yet appearing in Gemini shopping results, the underlying product data work is still worthwhile, but the urgency on UCP-specific integration is lower.

The pattern is clearer than it looks. The decision is rarely a binary "is this for us". It is "when do we engage, and at what pace". The answer depends on category maturity, competitor movement and product data readiness.

What EU commerce leaders should do now

The right response is not panic. The right response is preparation. UCP does not mean every brand needs to build a full agentic commerce integration tomorrow. But it does mean commerce leaders should start assessing whether their Google commerce stack is ready: Merchant Center, product data, checkout eligibility and governance.

Start with your most commercially important products or categories.

1. Audit Gemini visibility for your brand. Test whether Gemini understands your products, attributes, use cases and differentiators. A quick first test takes ten minutes. Open the Gemini app and run three to five category queries written the way one of your customers would write them. Note where your brand appears, what is said about it, and where competitor substitution happens. If your brand is missing or described in language that does not match how you would describe yourselves, you have a visibility gap worth diagnosing properly. Look beyond presence: check for misrepresentation, omissions, weak product matching and competitor substitution.
2. Audit Merchant Center and product data quality. Review titles, descriptions, attributes, images, variants, identifiers, pricing, stock, shipping, returns and warnings inside Merchant Center. Pay particular attention to mismatches between your website, your product feed, retailer listings and structured data. Strong Merchant Center data is the foundation for Gemini visibility and UCP eligibility.
3. Audit data integrity and claims. Identify claims that are vague, inconsistent or unsupported across your product detail pages, feed and structured data. This is especially important for premium, regulated, sustainability-led or technical products, where accuracy is part of trust.
4. Audit UCP commerce capability. Ask whether your systems could support reliable checkout-session logic, product ID mapping, payment handoff, shipping calculation, policy display and order lifecycle updates as documented in Google's UCP native checkout guide.
5. Define governance rules. Decide what Google's AI surfaces are allowed to access, recommend, repeat, initiate or hand off on behalf of your brand. Assign ownership across commerce, marketing, legal, technology and customer experience.

Other AI commerce systems, including ChatGPT, Perplexity and Claude, use different mechanisms, and brands will eventually need to understand those too. But for Universal Cart and UCP, the immediate readiness question starts with Google's stack: Merchant Center, product data, checkout eligibility and governance.

A practical readiness lens

The five actions above map into four broader readiness layers: can your brand be understood, trusted, technically selected and safely transacted with inside Google’s AI-mediated commerce stack?

This is the work now. Not because agentic commerce has fully arrived in Europe, but because the infrastructure is being built.

The commercial bottom line

Universal Cart makes the UCP story easier to understand. It shows what happens when AI shopping is not just a chatbot, a search result or a product recommendation, but a persistent layer that can carry customer intent across surfaces.

The website still matters. Brand still matters. SEO and performance marketing still matter. None of this is going away. What is changing is what they need to work alongside. The next layer of competitiveness, starting with Google’s AI commerce stack, will depend on whether AI systems can understand your products, trust your data, respect your rules and move customers safely towards purchase.

For EU commerce leaders, UCP should not be treated as a technical curiosity. It should be treated as an early operating signal.

The brands that prepare will not be the ones shouting loudest about AI. They will be the ones easiest for AI systems to understand, safest to recommend and most reliable to transact with.