Universal Cart is live, and Target is selling through it. Here is what brands need to do now.

Google's Universal Cart, the intelligent cross-merchant cart unveiled at I/O in May, has been seen operating in real accounts, and Target has confirmed it is now selling directly through Google's checkout. For brands selling across Google's surfaces, the question is no longer whether to prepare for agentic commerce, but whether your catalogue is in a state that lets you participate in it at all.

What is actually live

Universal Cart has gone from "coming this summer" to functioning in the wild. The cart works across merchants and across Google's own surfaces, so a shopper can add items while browsing Search, chatting in Gemini, watching YouTube or reading Gmail, and have them land in a single persistent cart. Once an item is added, the cart works in the background, finding deals, tracking price drops, surfacing price history and alerting on restocks. The significant part sits underneath that convenience. The cart reasons about the basket itself, checking compatibility across items even when they come from different retailers, flagging incompatibilities without being asked and suggesting alternatives. Because it is built on Google Wallet, it also factors in payment perks, loyalty status and merchant offers. Checkout runs through Google Pay in a few taps, and the brand stays the merchant of record throughout.

Target is the live proof. In a fact sheet published this month, Target confirmed that shoppers can now browse product listings and buy items directly from Target inside Google AI Search, including AI Mode, and the Gemini app, with the capability powered by the Universal Commerce Protocol that Target co-developed alongside Google. The detail in Target's own description shows what the live experience looks like from the brand side.

Target's walkthrough has a shopper open Gemini or Search and ask a conversational question, the example being a request for a trendy vacation bag to wear with light-washed jeans and brown boots that needs to fit a water bottle and a sweater. Google returns options from across Target's assortment, the shopper picks one, hits Buy, signs into their Target account and confirms. They can also transfer the cart to Target's own site to finish there. Payment runs through Google Pay with a major card or the Target Circle Card, promotions can be applied at checkout, and Circle members earn rewards and benefits in the flow. Target notes that multi-item purchasing in a single transaction on Google is coming soon, which lines up directly with what Universal Cart is built to do.

Two figures from Target's fact sheet frame why they are moving early. AI-driven traffic to Target rose 2,000% in the first quarter compared with the same period last year, against a roughly 400% rise in AI-driven traffic to retail sites overall. Target also positions itself as the first mass retailer with shopping experiences live across Google, Microsoft Copilot and OpenAI's ChatGPT. A major retailer is treating conversational surfaces as a primary channel rather than an experiment, and structuring its catalogue and loyalty programme to work inside them.

What this changes for brands

The conversational query in Target's example is the whole point. A shopper asking for a bag that fits a water bottle and a sweater, to wear with a specific outfit, is not searching on keywords. They are describing a need, and the agent has to match products against that need using structured attributes. Capacity, dimensions, materials, colour, style and occasion all have to be present and accurate in the feed, or the product cannot be matched.

This is where the cart's reasoning layer becomes decisive. When Universal Cart checks compatibility, fit, materials, ingredients, dimensions or care instructions, it is reading those fields directly from Merchant Center. Products with incomplete or inconsistent attributes get filtered out of the cart's reasoning before they ever reach the shopper. Because the cart reasons across the entire basket rather than a single product, this applies to every SKU in the catalogue, not just the bestsellers. A single missing field on a mid-tail product can be enough to disqualify it from a multi-item combination.

There is a second shift worth naming. The personalisation that makes the cart useful runs on data the brand never sees. Universal Cart draws on cart additions across Google's surfaces, purchase history, and cards, loyalty and offers through Wallet, and Google understands the shopper in far higher resolution than the brand making the sale does. Closeness to the customer used to be something brands built deliberately through email lists, accounts, apps and loyalty programmes. In an agentic environment, a large part of that closeness is now rented from the platform. Target's response, linking Circle accounts directly into the Google flow so members keep their benefits, is exactly the kind of move that protects the relationship rather than surrendering it.

What brands should be doing to win in Google's agentic shopping

There are two jobs. Make the product data good enough for the cart to reason in your favour, and get connected to UCP so that reasoning ends in a sale on Google's surfaces. The data work is organised around the 5 Cs of the agentic shelf, the framework we developed with the Digital Shelf Institute: Completeness, Context, Citations, Correctness and Customer Acquisition. The connection work is the UCP integration. Run them in parallel.

Get the product data right: the 5 Cs

The brands that win here will treat product data as infrastructure rather than as marketing content. Each of Google's new capabilities maps onto one or more of the five Cs.

Completeness: get every field populated, starting with the ones the cart reads

The cart cannot reason about a field that is blank. Work through the catalogue in this order.

First, pull a coverage report from Merchant Center and find every SKU with missing or "not specified" values on the attributes the reasoning layer uses: dimensions, capacity, materials, fit and sizing, weight, ingredients, compatibility and care instructions. These are no longer optional metadata. Treat a blank one the way you would treat a blank price.

Second, run the fill across the whole catalogue, not the top sellers. Mid-tail and long-tail SKUs need the same depth because one missing field on a single item can disqualify a multi-product combination from the basket. A realistic target is full population of the core attribute set on 100% of active SKUs.

Third, populate Google's new conversational attributes in Merchant Center, including the fields for common product questions and for compatible accessories and substitutes. Build the product-question entries from the questions shoppers actually ask, such as "is this dishwasher safe" or "will this fit a king bed", rather than restating the spec sheet.

Context: write attributes around the situations shoppers describe, not demographics

The agent matches a described need, so the work is making each product legible against real scenarios.

Separate what a product is from what it does, and connect the features to outcomes. A bag is not just "20 litres"; it is "fits a 15-inch laptop, a water bottle and a folded jacket". Map your top products against the conversational queries you expect, the way Target's example shopper asks for a bag to wear with light-washed jeans and brown boots, then check each attribute the query depends on is present and specific.

Replace vague use-case language with concrete situations. "Smart casual" is weaker than "suitable for a first day in a corporate office". "All-weather" is weaker than "rated for sub-zero commuting".

State limitations explicitly. Products that say what they are not for, such as "not suitable for children under three" or "not compatible with gas hobs", perform better because they let the agent rule the product in or out cleanly instead of guessing.

Citations: make your brand the credible source agents default to

When several products meet the same criteria, agents favour sources they recognise. Build that recognition deliberately.

Align the same product facts across your site, Merchant Center and every retailer feed, because a discrepancy between sources is read as a reason to discount you. Add and validate Organisation and Product schema on your own pages so the brand and product data are machine readable. Invest in owned depth, including detailed PDPs and buying guides that answer the questions agents are retrieving against. Then build earned presence in the reputable publications, marketplaces and structured sources that feed retrieval, and audit where your brand is currently being cited, where it is missing and how it is represented, so the gaps become a worklist rather than a guess.

Correctness: lock down data consistency and identifier integrity

This is the failure mode that silently removes products from recommendation and breaks checkout.

Reconcile attributes across your website, Merchant Center and retailer feeds and resolve every conflict, because inconsistency tells the agent the data cannot be trusted. Treat the feed as live infrastructure, with pricing, availability and identifiers updated continuously rather than cached or batched. Above all, audit identifier integrity. In Google's UCP model the agent passes a variant ID into checkout as a line item, so every variant ID, SKU and GTIN must be stable, unique and correctly mapped from discovery through to the buy button. A missing or mismatched ID does not degrade the experience, it stops the transaction. Run a recurring check that every purchasable variant resolves to one correct identifier across every system.

Customer Acquisition: protect the direct relationship the platform is renting back to you

The personalisation runs on data you never see, so the advantage goes to brands that secure their own.

Treat shoppers who choose to engage with you directly as a self-selected, high-value segment and invest harder in first-party data and direct relationships. Push owned data assets upstream rather than leaving them in a CRM no agent can read: feed loyalty status, purchase history and preferences into clean attribute structures, retailer integrations and signed-in experiences. Target linking Circle accounts into the Google flow, so members keep their rewards and benefits while buying through Gemini, is the template to copy. Then plan budget around the asymmetry rather than assuming you still own the relationship, because a plan that assumes otherwise will misallocate spend.

Get connected: enable UCP

UCP is the mechanism that turns the data work into transactions on Google's AI surfaces. It is the open standard, co-developed by Google with retailers including Target and Walmart, that lets agents and merchant systems speak a common language across discovery, checkout and post-purchase. Enabling it puts a checkout button on your eligible product listings in AI Mode in Search and the Gemini app, so shoppers buy in the flow while you remain the merchant of record.

Access is currently limited to selected merchants, and availability is rolling out in the US first, with Canada and Australia in the coming months and the UK to follow. The path to going live runs as follows.

First, confirm you meet the baseline. You need an active Google Merchant Center account, products eligible for checkout, and structured, accurate product data, which is where the 5 Cs work pays off directly.

Second, prepare the Merchant Center account. Configure shipping and returns, and validate the product feed so products can be discovered and bought on Google's surfaces. Incomplete shipping or returns configuration will hold up approval.

Third, join the UCP waitlist through Google's merchant interest form to start the implementation conversation. Integration must be approved by Google before you can go live.

Fourth, complete the technical integration. Publish your UCP profile so Google can discover your capabilities, payment handlers and public keys for signature verification, then implement the three core REST endpoints for creating, updating and completing a checkout session. Decide between guest checkout, which is the default and needs no extra steps, and account-linked checkout, which uses OAuth 2.0 to sync user profiles and is what lets loyalty benefits carry through, as Target does with Circle. Finally, wire up order-status webhooks so order updates push back to Google.

Where this leaves us

The cart is live, a major retailer is transacting through it, and the protocol behind it is expanding across markets and verticals on a monthly cadence. The cost of poor product data is now immediate and measurable, because the reasoning layer filters on it in real time. This is the early-mover window, and the brands that get their attribute coverage, context, trust signals and data consistency right now will be the ones the cart reasons in favour of, in the same way that early discipline on Amazon listings compounded into lasting advantage.

At Azoma, we help brands structure product data and optimise for discovery across the full set of AI shopping surfaces, organised around the 5 Cs of the agentic shelf. If you are thinking about how your brand shows up in Gemini and beyond, get in touch.