Google Universal Cart: When AI Becomes the Shopper, Who Controls the Sale?

Executive Summary

• Fact: At Google I/O 2026, Google launched Universal Cart — a persistent AI-powered shopping cart working across Search, Gemini, YouTube, and Gmail, with Gemini models monitoring prices and deals continuously in the background.

• Impact: The Agent Payments Protocol (AP2) will let AI agents complete purchases autonomously within user-set guardrails. The Universal Commerce Protocol (UCP) already counts Walmart, Target, Shopify, Visa, Zalando, Stripe, and 20+ major commerce players as signatories.

• Surprise: Branded searches in AI-powered surfaces already show an 18% higher click-through rate than generic queries — meaning catalog quality and brand recognition together determine who survives when the AI becomes the shopper.

Shopping has always had a funnel. Awareness, consideration, decision, purchase. For decades, that funnel ran through Google’s blue links. Then came comparison tools, product listings, AI Overviews. On Tuesday at Google I/O 2026, the funnel didn’t just evolve — it got a new manager. And that manager doesn’t scroll.

A Cart That Never Closes

Universal Cart is, on the surface, a deceptively simple idea: one persistent shopping cart that follows you across every Google surface. You’re browsing Search and add a wireless speaker. Later you watch a YouTube review of that same model. Then a Gmail newsletter from a retailer mentions a sale. Google is already tracking all of it.

What makes this structurally different from any saved-items list you’ve used before is what happens the moment a product enters the cart. Gemini models start working in the background — monitoring price drops, flagging stock alerts, surfacing price history — continuously, without the user doing a thing. The phone can be locked. The browser tab closed. The agent keeps working.

Then there is the Agent Payments Protocol, or AP2. Google announced at I/O that it is building the infrastructure for AI agents to complete purchases autonomously. A user sets constraints — a spending limit, preferred brands, specific product criteria — and when conditions are met, the agent buys. No confirmation screen. No browsing session. Google has confirmed AP2 will roll out to Gemini products in the coming months, starting with Gemini Spark.

There is a word for this moment. Agentic commerce. And it has arrived faster than most retail strategists expected.

When Walmart and Visa Sign the Same Protocol

Here is what gets lost in the I/O announcement noise: Universal Cart is not a Google product deployed in isolation. The Universal Commerce Protocol underpinning it was co-developed with Shopify, Etsy, Wayfair, Target, and Walmart. It carries endorsements from Adyen, American Express, Best Buy, Flipkart, Macy’s, Mastercard, Stripe, The Home Depot, Visa, and Zalando — more than 20 major commerce and payments players in total.

When a protocol is co-signed by both the infrastructure (Visa, Mastercard, Stripe, Adyen) and the retail layer (Walmart, Target, Home Depot, Macy’s) at the same moment, that is not a product launch. That is a standard being set.

What we’re seeing at Epinium is brands asking whether they need to prepare for agentic shopping. The answer, after this week, is no longer a strategic question. It is an operational one.

The contrarian perspective deserves airtime, though. Amazon’s Buy for Me and Walmart Sparky are making the same bet simultaneously — three major ecosystems racing to become the default AI purchase layer. Brands that integrate with one protocol will not automatically be present on the others. The near-term reality is fragmentation: separate data feeds, separate catalog requirements, separate agent behaviors. That creates real operational overhead for any team managing listings across channels.

Epinium data

Across more than 300 brand catalogs we manage across Google Shopping and Amazon, products with full attribute enrichment consistently outperform basic listings by 2.4× in click-through rates on AI-curated surfaces. As Universal Cart shifts purchase decisions to before the user actively searches, catalog completeness stops being a nice-to-have. It becomes the entry ticket.

Is your brand ready for the agentic shopping era? See how Transform helps brands build the strategy and infrastructure this shift demands → ✓ 30 min  ✓ No cost  ✓ Dedicated AI Director

The Invisible Brand Problem Nobody Is Talking About

Universal Cart creates a question that isn’t getting enough attention: what happens to brands whose product data doesn’t surface in the UCP graph?

The short answer is they don’t get added to carts. AP2 agents pull from structured product data exposed through the protocol — primarily via Google Merchant Center feeds. Incomplete attributes, mismatched GTINs, stale pricing: the agent skips those listings. Not out of editorial judgement. Because they don’t appear as valid matches against the user’s criteria.

Research from Amsive found that branded searches in AI Overview surfaces already generate an 18% higher click-through rate than generic queries. That gap carries an important signal: brand equity still matters in the AI era, but only when your structured data allows the brand to surface. Strong brand recognition and poor catalog hygiene cancel each other out.

The practical implication is narrow and concrete. Google confirmed Universal Cart is expanding to Canada and Australia in the coming months, with the UK following. Brands with international catalog operations have a real window — not a wide one — to clean their Google Merchant Center feeds before the protocol becomes baseline infrastructure rather than a differentiator.

For brand managers and catalog teams, three actions matter right now: audit feed completeness against Google’s structured attribute requirements, verify GTIN accuracy at the SKU level, and confirm that pricing and inventory signals are updating in near-real time. None of this is glamorous. But it determines whether an AI agent adds your product to a user’s cart or scrolls invisibly past it — in a session your analytics dashboard will never record.

For broader context on where this trajectory leads, see Agentic Commerce $1 Trillion by 2030: Is Your Catalog Ready? — and what Amazon’s Alexa-for-shopping shift tells us about the direction every major platform is moving.

Frequently Asked Questions

How is Google Universal Cart different from a standard saved-items or wishlist feature?

A wishlist is static — you add things and they sit. Universal Cart is active: Gemini models monitor prices continuously, send deal alerts, flag back-in-stock events, and — via AP2 — will eventually execute purchases autonomously within user-defined limits. It’s the difference between a notepad and a personal buyer who works around the clock. That distinction fundamentally changes how brand discovery and conversion interact.

Does a brand need to be a large retailer to appear in Universal Cart?

No. Integration happens through the merchant’s structured product feed, primarily via Google Merchant Center. Brands of any size can be surfaced if their catalog data is accurate, complete, and attribute-rich. The competitive disadvantage isn’t scale — it’s data quality. A niche brand with clean, enriched feed data can appear alongside a major retailer’s listing when the agent’s criteria match. The playing field is technically open; the execution gap is where smaller brands tend to fall behind.

What guardrails exist to prevent AP2 agents from making unauthorized purchases?

AP2 requires users to configure explicit spending limits, brand preferences, and product criteria before an agent can act. Google describes the mechanism as a “tamper-proof digital mandate” — a verifiable link between user, merchant, and payment processor using privacy-preserving technology. The agent can only complete a purchase when all pre-set conditions are met. Critically, promotional discounts only influence an AP2 agent if they appear in the structured product data feed. A banner ad or email promotion the agent never processes will have zero effect on whether it buys.

When should a brand hold off on integrating with Universal Commerce Protocol?

If the catalog isn’t clean, integrating early could actively hurt performance. An agent operating on messy feed data — duplicate SKUs, incorrect pricing, missing category attributes — will produce irrelevant results, which erodes trust with the end user and may train the agent to deprioritize the brand’s listings. Cleaning the feed first, then integrating, is the more defensible sequence. A rushed integration with bad data is a worse outcome than no integration at all.

How does Universal Cart coexist with Amazon’s Buy for Me and Walmart Sparky?

Each platform’s agentic shopping layer operates independently. UCP is a Google-published protocol not shared with Amazon or Walmart. Brands seeking presence across all three AI purchase layers will need well-structured feeds on each system: Google Merchant Center for Universal Cart, Amazon’s catalog infrastructure for Buy for Me, and Walmart Connect for Sparky. Multi-platform catalog management — centralizing enrichment and keeping feeds synchronized — will be the operational capability that separates brands that scale efficiently from those that drown in manual feed maintenance as the protocols mature.

The shopping session of the future begins before the consumer actively shops. An AI agent, operating on background context and pre-set preferences, filters the product universe before a human finger ever touches a screen. Brands that invest now in catalog quality, structured data, and cross-platform feed management are not preparing for some distant AI future. They are catching up to an infrastructure that 20-plus global commerce leaders signed onto last week in Mountain View.

Transform by Epinium Ready to prepare your brand for the agentic commerce era? Epinium’s Transform practice builds the catalog infrastructure, feed strategy, and AI readiness plan that Universal Cart demands — with a dedicated AI Director and a diagnosis tailored to your category. Book your free 30-min diagnosis → 30 min · No cost · Personalized diagnosis