Google Wants to Own the Checkout Button

There's a moment in every platform company's life where it stops being infrastructure and starts being a landlord. Google just crossed that line, and most people are too busy admiring the product to notice.

At Google I/O on May 19 2026, the company unveiled Universal Cart: a persistent, AI-powered shopping cart that follows you across Search, Gemini, YouTube, and Gmail. Add a product while watching a YouTube review, check out later from your Gmail inbox. No new tabs, no re-entering your card, no navigating unfamiliar checkout flows. Just a frictionless tunnel from "I want this" to "purchased," entirely inside Google's walls.

It's genuinely impressive. It's also one of the most strategically significant moves Google has made in a decade.

---

What Is Actually Being Built

Universal Cart is the consumer-facing tip of a much larger agentic commerce infrastructure. Underneath it sit three interlocking pieces that, taken together, amount to a complete rebuild of the purchase funnel.

The first is the Universal Commerce Protocol, announced in January 2026 and co-developed with Shopify, Walmart, Target, Etsy, and Wayfair. It's an open standard that gives AI agents a shared grammar for the entire shopping lifecycle: product discovery, cart management, checkout, tax, shipping, and post-purchase support. Think of it the way HTTP works for web pages: any compliant system becomes interoperable with any other, with no custom integrations required.

The second is the Agent Payments Protocol, the execution layer that lets AI agents complete purchases autonomously within guardrails the user sets: spending caps, approved brands, specific categories. There is a permanent audit trail for every transaction, which matters both for dispute resolution and for the regulatory audiences Google is clearly already anticipating.

The third, and least discussed, piece is the Shopping Graph: a catalog of over 60 billion product listings indexed from across the open web. That number matters because Amazon has blocked rival AI platforms from accessing its catalog entirely. When OpenAI's shopping agent looks for a product, it cannot surface anything from Amazon. Google's can. That asymmetry is significant, and it's largely absent from the coverage.

Put all three together and you are not looking at a checkout feature. You are looking at a complete rebuild of the purchase funnel as a Google-native experience.

---

The Protocol War Nobody Is Covering Properly

Universal Cart is built on a standards battle that Google has effectively already won. How it won matters more than the product itself.

In September 2025, OpenAI and Stripe launched a rival standard: the Agentic Commerce Protocol, powering ChatGPT's Instant Checkout. For a window, it looked formidable: Stripe's payment infrastructure behind it, OpenAI's large user base in front. Then it quietly stumbled. By March 2026, OpenAI was walking back the implementation. The original version wasn't flexible enough for the messy reality of production retail: regional tax rules, inventory edge cases, loyalty integrations.

	Universal Commerce Protocol	Agentic Commerce Protocol
Led by	Google	OpenAI + Stripe
Launched	January 2026	September 2025
Coalition	Shopify, Walmart, Target, Amazon, Meta, Microsoft, Salesforce, Stripe	OpenAI, Stripe
Current status	Active governance council	Partial rollback (March 2026)
Stripe's role	Governance member	Co-creator

Google, meanwhile, was playing coalition politics. By late April 2026, the Universal Commerce Protocol Technology Council had absorbed Amazon, Meta, Microsoft, Salesforce, and, critically, Stripe itself. The same Stripe that co-built the competing protocol is now a governance member of the body that steers this one.

The reason Google won is worth stating plainly: it can afford to give the protocol away for free because it owns the infrastructure beneath it. Every merchant that adopts the standard sends more queries through the Shopping Graph. Every transaction runs on Gemini's infrastructure. The protocol is open. The moat is not.

---

The Problem For Merchants

Google's official framing is reassuring: merchants remain the Merchant of Record, they own the customer relationship, they own the data. Technically, all of that is true, and it was deliberately designed that way, partly to sidestep the financial regulations that come with taking custodial ownership of transactions.

But "you own the relationship" means something quite different when the discovery, comparison, cart, and checkout all happen inside a Google surface. The customer never visits your site. They never see your brand story. They don't join your email list. Instead, they're interacting with an AI that ranked your product among several options, and the ranking logic is not visible to you.

This is the Amazon problem in newer clothes. Amazon sellers technically own their products and pricing. Any seller who has watched their visibility collapse after an unexplained algorithm change knows how hollow "ownership" becomes when the platform controls distribution.

The concern is not theoretical. Google Shopping was already the subject of a landmark European Commission antitrust ruling in 2017, upheld on appeal in 2024, specifically for self-preferencing its own service while demoting competitors. Universal Cart extends that structural pattern into full agentic transactions.

---

The Consumer Trade-off

For shoppers, the pitch is genuinely compelling. Online checkout is still surprisingly fragmented for something that has existed for 30 years. Universal Cart eliminates most of that friction: passive price tracking, automatic restock alerts, loyalty points that apply without you having to think about them.

But convenience has a cost that surfaces later. When an AI agent is monitoring prices, surfacing compatible products, and nudging you toward checkout, who is it optimizing for? It might be the highest-bidding merchant. It might be Google's revenue goals. You will not know, because the agent does not explain its reasoning.

A Bain and Company survey from late 2025 found that only 16 percent of shoppers trust a platform like Google to carry out end-to-end purchasing on their behalf, compared to 25 percent who trust retailer-owned tools. People are comfortable with Google finding things. They are less comfortable with Google buying things for them. Whether that gap closes depends on how transparent Google makes the agent's reasoning, and its track record on algorithmic transparency is not encouraging.

---

What Builders and Merchants Should Do

Treat protocol support as a distribution decision, not a loyalty test. You need the Universal Commerce Protocol for Google's surfaces and the Agentic Commerce Protocol for ChatGPT. Stripe's decision to join both governance bodies is the clearest signal available that the payments layer will stay protocol-agnostic. Build your payments stack on that layer regardless of which standard wins.

Merchant Center optimization is now as strategically important as search engine optimization was in 2015. Product feed quality, real-time inventory accuracy, and structured attributes determine your visibility across AI surfaces. That is where investment should go.

And keep building your direct channel: email list, owned app, loyalty program. Not because agentic commerce will fail, but because concentrated distribution is a business risk. The merchants who will regret this decade most are the ones who let their direct customer relationships atrophy while chasing platform visibility.

---

The Bigger Picture

Step back and what you are watching is the internet's commerce layer being renegotiated. Search centralized discovery. Social media centralized attention. Agentic AI is now centralizing action, and the entity that controls the action layer captures value at a scale that makes the previous two shifts look incremental.

Google's bet is that it can become the neutral infrastructure everyone builds on: open enough to attract the coalition that makes the standard real, central enough to compound the data advantage, and necessary enough that opting out means losing a distribution channel you cannot easily replace.

Whether that resolves as revolutionary convenience or as a new generation of platform capture depends on which side of the counter you are standing on. If you are standing on both, as most people in this industry are, the honest answer is that it is both, and the ratio between the two is being decided right now.

---

Key Takeaways

• Google's Universal Cart is built on three infrastructure layers: the Universal Commerce Protocol, Agent Payments Protocol, and a 60-billion-listing Shopping Graph. It's not a checkout feature. It's a purchase funnel replacement.
• The protocol battle with OpenAI's Agentic Commerce Protocol is effectively over. Stripe now sits on both governance bodies. UCP has the coalition.
• Merchants technically own the customer relationship, but visibility depends on an AI ranking algorithm they cannot audit. It's the same structural position Amazon sellers are already in.
• Only 16% of shoppers trust a platform like Google to complete purchases on their behalf, versus 25% for retailer-owned tools (Bain, 2025).
• Practical response: support both protocols, prioritize Merchant Center feed quality, and keep building direct customer channels.