In January 2026, Google unveiled the Universal Commerce Protocol, or UCP for short. Built alongside Shopify, Etsy, Wayfair, Target, and Walmart, and endorsed by over 20 companies including Visa, Mastercard, and Stripe, UCP gives AI agents a shared language for handling the full shopping journey.
Then on March 19, Google announced a major expansion: three new capabilities that move UCP from basic single-item checkout toward a complete shopping experience inside AI surfaces.
If you sell online, you need to understand what these capabilities do and how to prepare for them.
What UCP Does (Without the Jargon)
Think of UCP as a menu your store publishes for AI agents. Instead of a customer browsing your website, an AI agent reads your menu, understands what you sell, checks what's in stock, builds an order, and completes the purchase, all inside Google's AI Mode or Gemini.
Your store tells the AI what it can do. The AI figures out what it can handle. They agree on the overlap, and the transaction happens.
You stay the merchant of record. You keep the customer data. You control the relationship after the sale. But the browsing-and-buying happens on Google's surface, not yours.
The Building Blocks: UCP Capabilities Explained
UCP is modular. You pick which capabilities you want to support. Each one handles a different part of the shopping journey. As of March 2026, UCP checkout on Google's surfaces is limited to eligible US merchants via a waitlist, with global expansion planned over the coming months. The capabilities below are at different stages of rollout.
- Checkout (live since January 2026)
This is the foundation. Checkout lets an AI agent create an order session, add items, calculate taxes, handle shipping options, and process payment. When a shopper asks Gemini to find them a carry-on suitcase and decides to buy one, this capability powers the transaction.
Wayfair and Etsy were among the first retailers to go live with UCP checkout in AI Mode from February 2026. Shopify, Target, and Walmart integrations have been confirmed but don't have public launch dates yet. The buy button is already appearing on eligible product listings across Google's AI surfaces.
Checkout also has built-in extensions for discounts (so agents can apply promo codes) and fulfilment (so they can offer different shipping methods, pickup, or delivery windows). - Order Management (live since January 2026)
Once a purchase happens, the conversation doesn't end. Order Management keeps the AI agent informed about what happens after checkout: shipping updates, delivery confirmations, returns, and refunds. A shopper can ask their AI assistant "where's my order?" and get a real answer pulled from your system. - Identity Linking (live since January 2026, highlighted in March update)
This one matters for anyone running a loyalty programme. Identity Linking lets shoppers connect their existing store account to UCP using a standard login flow. Once linked, the AI agent knows the shopper's membership tier, their loyalty pricing, their saved payment method, and their shipping preferences.
The practical effect: a Walmart+ subscriber shopping through AI Mode gets the same free shipping they'd get on walmart.com. An Ulta Beauty Rewards member still earns their points. Without this, loyal customers have a reason to leave AI Mode and go buy from your site directly, which is good for site traffic but bad for conversion in the agentic channel. - Cart (new, announced March 19, 2026, draft specification)
Until March, UCP could only handle single-item checkouts. The new Cart capability changes that. AI agents can now save or add multiple items to a basket from a single store, the same way a regular shopper would.
Carts are designed for browsing and building. A shopper can tell the AI to put together an outfit or gather supplies for a project, and the agent assembles the cart over the course of the conversation. When the shopper is ready, the cart converts to a checkout session.
This is a draft specification for now, meaning it could still change based on community feedback. But the direction is clear: Google wants multi-item shopping to work inside AI surfaces. - Catalog (new, announced March 19, 2026, draft specification)
This is the capability that caught the most attention in the March update. Catalog lets AI agents pull real-time product details directly from your inventory, including variants, current pricing, and stock levels.
Before Catalog, agents relied on your Merchant Center product feed for discovery. Feeds are updated periodically. Catalog adds a live layer on top of that, so the agent can check whether the medium blue version is still in stock right now, at the current price, before it recommends it to a shopper.
For merchants, this raises the stakes on inventory accuracy. If your systems say something is available but it's not, the agent will attempt the transaction and fail. Each failure erodes your reliability score, and the AI starts recommending you less often.
For merchants looking to support Catalog, Feedoptimise can serve as the real-time data layer for catalog retrieval with UCP. Our platform already integrates with Shopify, WooCommerce, Magento, and other major platforms, so once the Catalog specification is finalised, from day one, we can connect your live stock and pricing data to meet UCP's requirements without needing to build a custom integration.
What's Coming Next
Google's roadmap for UCP mentions several features still in development: multi-item carts across multiple stores, account creation during checkout, subscription management, and deeper post-purchase support for tracking and returns.
Google is also simplifying the onboarding process through Merchant Center, aiming to bring in retailers of all sizes over the coming months.
How to Prepare: Practical Steps
- Get Your Product Feed in Shape
Your Merchant Center feed is still the primary way agents discover your products. A feed with sparse data means the AI skips you.
Titles should be 30+ characters with brand, model, and key attributes. Descriptions should be 500+ characters and answer common buyer questions. Include GTINs where available. Add at least three images per product. Set precise availability states with handling times, not just "in stock" or "out of stock."
Merchants with near-complete attribute data are seeing 3-4x higher visibility in AI recommendations. - Enable UCP Eligibility
Add the native_commerce attribute to your product feed. Products without this attribute won't show the checkout button on AI surfaces. Complete your return policies at every account level. Add customer support information. Check Google's product restrictions list against your catalogue. - Prepare for Catalog Access
If you plan to support the Catalog capability, your inventory systems need to provide accurate, real-time data. The agent will be querying your stock levels and pricing live, not relying on a feed that refreshes every few hours. Price mismatches between your feed and your actual site are already a problem for standard Shopping ads. Under Catalog, they become a bigger one. - Think About Identity Linking
If you run a loyalty programme or membership pricing, decide whether you want those benefits to carry over into agentic transactions. The upside: loyal customers get a seamless experience and you avoid losing them to a "I'll just buy it on the website" moment. The downside: one more integration to maintain, and you're extending your loyalty economics to a channel you don't control. - Protect Your Margins
AI agents optimise for conversion probability. In example: They don't know which products make you money and which ones have a 3% margin with a 40% return rate. Feed profit signals into your campaigns through custom labels. Segment by margin tier. Use value-based bidding. If you let agents sell whatever converts, you'll fill your order book with the wrong products. - Join the Pilots
Google is running several early programmes worth exploring: Business Agent (a branded AI sales associate in Google Search), Direct Offers (exclusive discounts triggered by high purchase intent in AI Mode), and the Conversational Attributes Pilot (new Merchant Center fields for question-and-answer data, compatible accessories, and product substitutes). Early participants shape the tools and build an advantage before general availability.
How we can help
Everything in this article comes back to one thing: product data feed quality. Your titles, descriptions, attributes, stock levels, pricing, GTINs, images, shipping details, margin labels. Getting all of that right across hundreds or thousands of SKUs, and keeping it right as UCP and its capabilities evolve, is a serious operational challenge.
That's what we do at Feedoptimise. Our platform and managed service handle the full feed lifecycle, from pulling your raw product data out of Shopify, WooCommerce, Magento, BigCommerce, Salesforce or any other platform, through to enriching, optimising, and syncing it across every channel you sell on.
A few things make our platform particularly suited for UCP readiness. Our AI feed enrichment tools can rewrite titles and descriptions to meet the longer, more detailed formats that AI agents need for conversational discovery. Our rule engine lets you set up margin-based exclusions and custom labels (season, margin bucket, best-seller) so you control which products the AI picks up. Real-time stock sync keeps your availability data accurate, which matters even more once the Catalog capability goes live and agents start querying your inventory directly. And our managed service team handles the day-to-day feed operations, from setting up needed attributes (like the native_commerce attribute Google requires) to monitoring for errors and disapprovals before they damage your visibility.
If you're looking at the preparation steps in this article and thinking "I don't have the team or the time to do all of this," get in touch. We'll help you get your feeds UCP-ready.
To sum up
UCP is moving fast. In January, it could handle a single-item checkout. By March, it supports multi-item carts, live catalog queries, and loyalty integration. Commerce Inc, Salesforce, and Stripe are building support into their platforms. And we at Feedoptimise are working on supporting Catalog capabilities through our platform the very moment Google releases it from the draft stage.
The common thread across all of this: rich, accurate, continuously updated product data wins. The AI can't sell what it can't understand. Merchants who treat their product data feed as a strategic asset, not a compliance checkbox, will capture sales in a channel that some projections say could reach $9 trillion by 2030.
Start with your top products. Get the Google Merchant Center’s feed right. Enable the attribute across relevant items. Then decide which capabilities make sense for your business. The protocol is modular for a reason. You don't have to adopt everything at once.