Google wants your shopping cart to think for itself.

At Google I/O 2026, the company quietly did something pretty radical for e-commerce: it turned the humble cart icon into an AI-powered command center called Universal Cart, and tied it to a broader vision it keeps calling “agentic commerce.” If Google pulls this off, you won’t just search for products and click “buy” anymore – you’ll increasingly tell an AI what you want, and let it handle the mess of deals, compatibility, loyalty perks, and checkout in the background.

For now, Universal Cart looks like a feature. In reality, it’s Google declaring that the future of online shopping won’t live on store websites, but in AI agents that move across them.

Universal Cart: the cart that works everywhere

Think about how you shop today. You might Google a product, watch a few YouTube reviews, chat with an AI like Gemini to compare options, click through to a couple of retailers, add items to two or three different carts, and maybe lose track of one along the way. Universal Cart is Google’s attempt to snap all of that into a single, persistent shopping brain that follows you around Google’s ecosystem.

The new cart is built directly into Search and the Gemini app to start, with YouTube and Gmail support coming later. You’ll be able to drop products into it from across Google – while browsing Search results, chatting with Gemini, watching a creator’s video, or even reading a promo email. Instead of a dozen disconnected mini-carts on different sites, Universal Cart becomes your default shopping hub on Google.

Once an item lands there, the interesting part begins: the cart doesn’t just hold it, it works on it.

Deals, price history, and a background AI constantly watching your cart

Behind the scenes, Universal Cart is plugged into Google’s Shopping Graph – the company’s giant real-time dataset of products, sellers, and pricing that already tracks tens of billions of product listings. Google says people already shop across its surfaces more than a billion times a day, and the Shopping Graph has grown to more than 60 billion product listings, from big-box stores to small retailers.

Universal Cart uses that data plus Google’s Gemini models to quietly babysit your cart. Once you add something:

• It hunts for deals and price drops and can surface price history so you’re not guessing whether “40% off” is actually special.
• It can notify you when an item comes back in stock, so you don’t have to constantly refresh pages.
• It factors in your payment perks, loyalty programs, and merchant offers through its integration with Google Wallet, surfacing ways to save or earn points that you might have forgotten about.

The endgame is that the “cart” becomes less of a waiting room and more of an active agent looking for better outcomes on your behalf. Google is clearly betting that everyday shoppers won’t mind an AI quietly optimizing their purchases, as long as it saves them money and hassle.

Agentic shopping in practice: “build me a PC that works”

Where this starts looking less like a clever coupon tool and more like a different paradigm is in Universal Cart’s “intelligent reasoning.”

Google’s example is the classic PC-building headache: you assemble a list of components from different retailers – CPU, motherboard, RAM, GPU, case, power supply – and hope they all play nicely together. With Universal Cart, once those items are in your cart, the system can proactively scan for incompatibilities and flag things that won’t work, then suggest alternatives that do.

That’s the essence of agentic shopping: instead of the web saying “you can buy this,” it starts saying “given what you’re trying to do, here’s a setup that will actually work – and here’s how to make it cheaper.” The agent isn’t just retrieving information; it’s taking initiative based on your goal.

In a couple of years, it’s not hard to imagine this extending far beyond PCs:

• “Plan a week of dinners under $80 for a family of four, and stock my cart with what I need.”
• “I’m moving into a studio; give me the essentials – furniture, kitchen gear, cleaning supplies – and prioritize space-saving items.”
• “I want to train for a 10K; build a starter kit with shoes, clothes, a watch, and some recovery gear within my budget.”

Universal Cart is clearly designed to be the surface where these kinds of agent-driven tasks land. Gemini (and eventually Gemini Spark, Google’s 24/7 agentic assistant) becomes the conversational front-end; the cart is where those conversations are turned into actual products and purchases.

UCP: the common language that makes this possible

Universal Cart alone doesn’t explain how an AI agent can safely and reliably talk to thousands of different retailers. For that, Google has been building out something more foundational: the Universal Commerce Protocol, or UCP.

UCP is an open standard that defines a common way for AI agents and retailers to interact – covering things like catalog search, cart building, identity linking, checkout, and order management. Instead of every merchant having to build bespoke integrations for each AI platform, UCP gives them one shared “language” so agents can discover products, check availability, assemble carts, and complete orders in a predictable, auditable way.

Google co-developed UCP with major retail partners and has been steadily adding more tech companies and commerce platforms to the working group. The protocol is already powering new checkout experiences on Google in the US and is being expanded to Canada, Australia, and later the UK, with more surfaces like YouTube and verticals like hotel bookings and local food delivery on the way.

If Universal Cart is the shiny feature users see, UCP is the merchant-facing plumbing that convinces retailers they’re not just handing control of their customer relationships to a black box. The pitch here is: this is still your customer, your brand, your order – we’re just giving AI a standardized way to help them shop and pay.

From “pay with card” to AP2: how do you let an AI spend your money?

The other uncomfortable question in an agentic future is obvious: if an AI can click “buy” for you, what stops it from overspending or making the wrong purchase?

That’s where Google’s Agent Payments Protocol (AP2) comes in. AP2 is a separate open protocol Google developed with financial and tech partners to create a shared rulebook for payments initiated by agents, not humans clicking buttons.

In AP2, the core concept is something called a Mandate: a cryptographically signed, tamper-proof digital contract that encodes exactly what you’ve authorized an agent to do. A mandate might say, for example: “You can buy this specific product from these brands, up to this price, within this time window, using this payment method.” The agent can only complete the transaction if the conditions match that mandate.

This does a few important things:

• It gives you explicit control over how much freedom your AI has to spend on your behalf.
• It creates a verifiable trail all parties can see – you, the merchant, and the payment processor – which helps with disputes and returns.
• It makes it possible to scale agentic payments across different payment rails (cards, bank transfers, new methods) in a consistent way.

Google plans to start bringing AP2 into its products with Gemini Spark, its more persistent, agent-like assistant. The idea is that you’ll be able to say things like “if the price drops below $200, buy it for me,” and know there’s a formal, cryptographically enforced constraint behind that promise instead of vague trust in an algorithm.

What this means for retailers

For merchants, Universal Cart probably triggers mixed feelings.

On one hand, Google is promising a smoother path into the agentic world. If you support UCP and plug your catalog, pricing, and inventory into that standard in a structured way, you become “AI-ready” across any platform that speaks UCP – not just Google. Industry observers have already framed UCP as a signal that commerce is standardizing around agent-compatible APIs, and smart retailers are being urged to clean up their product data, enrich attributes like compatibility, sustainability, and usage constraints, and make pricing and availability “real-time enough” for agents to act on.

On the other hand, Universal Cart undeniably shifts power toward aggregators. If a shopper increasingly lives inside Google’s cart, the retailer’s website becomes just one possible execution layer – the place where the transaction gets finalized, but not necessarily where discovery or decision-making happens. Google has been careful to stress that the merchant remains the “merchant of record,” but the AI assistant increasingly owns the customer journey.

The upside: AI-driven tools in Google’s Merchant Center will give brands more visibility into how they appear in agentic search results and help them optimize product content and offers for these new surfaces. The downside: you’re competing to influence an AI intermediary that’s designed, in theory, to be ruthlessly shopper-centric.

Why this matters for shoppers

If you’re just trying to buy a pair of sneakers, some of this sounds abstract. The immediate wins from Universal Cart are more grounded:

• Less tab chaos: one unified cart across Search, Gemini, YouTube, and Gmail instead of fragmented carts on dozens of sites.
• Better timing: active tracking for price drops, restocks, and promos so you’re not missing deals or checking manually.
• Fewer mistakes: intelligent checks for things like incompatible PC parts or mismatched accessories, plus suggestions that actually fit your setup.
• Faster checkout: Google Pay-based flows and UCP-powered checkout that work with big retailers like Nike, Sephora, Target, Ulta Beauty, Walmart, Wayfair, and select Shopify brands such as Fenty and Steve Madden.

In the near term, the experience might feel like a smarter version of what you already know: the familiar Google-y surfaces, but with a cart that remembers more, notices more, and occasionally nudges you with “hey, you might want to know this before you buy.”

Zoom out a bit, though, and Universal Cart is part of a clear shift. Google’s search and shopping experiences are steadily moving from query-driven (“best running shoes under $150”) to intention-driven (“I want to start running 3 times a week, help me get set up”). In that world, the winner isn’t just the store with the lowest price or most ads – it’s the assistant that best understands what you’re actually trying to achieve and can orchestrate the entire journey, from research to purchase to follow-up.

Agentic commerce is still early. Protocols like UCP and AP2 are new, and not every retailer will jump in at the same pace. There will be hard questions about bias, transparency, and how much agency we’re comfortable giving to AI systems that also happen to be advertising platforms. But Universal Cart is a tangible, user-facing step toward that future – one you’ll start to notice in your Google results and Gemini chats this summer in the US, before it spreads to more services and regions.