For twenty-five years selling online had a stable geometry: the user arrives on your site, your page convinces them, your checkout takes the money. Every optimisation lever we know — copy, images, social proof, friction removal — assumes that geometry.
Agentic commerce puts it in question. If an assistant can compare, choose and complete the purchase for the user, the product page stops being where the decision is made.
What UCP is
The Universal Commerce Protocol is an open standard Google released in January 2026, designed to give a common language to transactions between consumer surfaces, businesses and payment providers. It was co-developed with Shopify, Etsy, Wayfair, Target and Walmart, and endorsed by more than twenty partners including Visa, Mastercard, Stripe, American Express, Best Buy and The Home Depot.
That list of backers is the most informative part of the announcement: when payment networks, ecommerce platforms and large retailers align on the same protocol, the signal is that the infrastructure gets built regardless of how quickly individual merchants adopt it.
Agentic checkout
Retailers that enable UCP can add a checkout button to eligible items inside AI Mode and Gemini. The feature lets an agent initiate and complete the checkout process on a shopper’s behalf. At the time of writing, UCP-powered checkout is available to eligible US-based merchants, with global expansion planned through 2026.
Universal Cart
Universal Cart collects products added across different retailers into one cart, so shoppers see everything in one place. The rollout started with Search and the Gemini app in the US, with YouTube and Gmail to follow. Announced brands for the experience include Nike, Sephora, Target, Walmart, Wayfair and Shopify merchants such as Fenty and Steve Madden.
The underlying change. This is not "another sales channel". It is the possibility that comparison between alternatives happens off your site, on data you supply, with the decision made by a system reading that data rather than looking at your page. Storefront quality matters less; data quality matters a great deal more.
What changes concretely for a merchant
| Dimension | Traditional commerce | Agentic commerce |
|---|---|---|
| Where the decision happens | On the product page | In the agent’s interface, on feed data |
| What persuades | Copy, images, social proof, urgency | Comparable attributes: price, availability, shipping, returns |
| Role of the feed | Feeds ads and listings | It is the product’s representation |
| Main risk | Low conversion rate | Not being eligible, or being represented badly |
| Customer data | Collected by the merchant | Mediated, with consequences for retention and remarketing |
What to do now, outside the US
Said plainly: at the time of writing, UCP checkout covers eligible US-based merchants, and global expansion is announced but not complete. No merchant outside the US needs to reorganise a technical roadmap around this news today.
That said, three lines of work make sense regardless, because they pay off with or without agentic commerce.
1. Completeness and reliability of product data
If an agent chooses on attributes, every missing attribute is one fewer chance of being selected. Materials, dimensions, compatibility, delivery times, returns policy: these are the fields an automated comparison is built on. It is the same work described in the article on the 2026 product data specification, with one more reason behind it.
2. Consistency between systems
Availability and price must match across site, feed and stock system, quickly. In a mediated checkout, a mismatch is not minor friction: it is a transaction failing inside somebody else’s interface, with the reputational cost that follows.
3. A customer relationship that does not depend on the point of purchase
This is the serious strategic question. If the purchase happens through an intermediary, the merchant risks losing direct contact. The countermeasures are the familiar ones, but they become more urgent: loyalty programmes with a real reason to join, first-party data collected under clean consent, post-purchase content that brings the relationship home. On keeping the measurement of that relationship intact, we wrote the Consent Mode v2 guide.
A realistic perspective
It is worth being honest about the unknowns. It is not yet clear how attribution will be measured when a purchase happens inside an agent’s interface, how brand positioning will be defended in an automated comparison where the brand is one attribute among many, or how many shoppers will genuinely adopt agentic checkout in categories where the choice is emotional rather than comparative.
What can be said with reasonable confidence is that enough actors have chosen this direction for it to become infrastructure. Anyone selling online would do well to treat their feed the way they treat their site: not as a technical file to export, but as the way the product exists for the rest of the world.
If you want to know how ready your catalogue is for this, the free audit starts from feed completeness, data consistency and the strength of your customer relationship. See also our ecommerce services.