For years, the SEO for ecommerce had a clear logic: appear in search results, get the click, take the user to the store, convert.
The funnel was predictable. What Google announced in January 2026 changes that logic quite fundamentally.
The Universal Commerce Protocol, or UCP, is a protocol that allows artificial intelligence agents to complete complete commercial transactions without the user having to leave the interface where they are.
Search, compare, add to cart, pay, track the order: all within the agent. The online store still exists, but the user may never visit it.
Before UCP, Google and other AI systems could show products in their results. They could recommend options. But to buy, the user had to click and finish the transaction on the seller's site.
UCP eliminates that step. It defines a standard protocol that covers six capabilities of the full business cycle:
The difference with previous protocols such as ACP, which focused mainly on the checkout and payment flow, is that UCP covers the entire cycle. From the moment the user discovers the product until they receive post-sales support.
And it's already working. On February 11, 2026, Vidhya Srinivasan, VP of Advertising & Commerce at Google, confirmed that Wayfair and Etsy are already operating with UCP, allowing direct purchases within Google's AI Mode.
The adoption of new web protocols is often slow. UCP might seem like another technology that you have to follow from afar before investing time in it. There are two reasons why that position is a mistake in this case.
The first is the speed of implementation. Google announced UCP in January and already has retailers operating with the protocol in February.
The published roadmap includes expansion to new verticals, multi-product carts, integration of loyalty programs and post-sales support over the coming months. This is not in the experimental phase, it is in active deployment.
The second is that preparing for UCP requires basic work that the team probably has pending anyway: full schema markup, updated Merchant Center, detailed product attributes.
None of that is work that is lost if UCP takes longer than expected. Everything improves current search and paid performance.
To understand what needs to be done, we must first understand how selection works in an agentic environment.
When a user asks an AI agent to recommend shoes for running on mixed surfaces with a budget of 150 euros, the agent does not do a traditional search.
It does what is called a fan-out: it breaks down the query into multiple intentions and attributes, consults structured sources, crosses availability, price, ratings and compatibility data, and generates a selection.
In that process, products that have detailed and well-structured attributes are more likely to be selected.
A product that has only a name, price and generic description competes at a disadvantage against one that specifies material, compatibility, size and color variants with specific names, frequently asked questions answered, and verifiable third-party reviews.
The color “purple” is less accurate than “dark purple”. But there are brands that go further and use proper names for their colors, such as “Wolf” for a specific gray tone in On sneakers. That granularity isn't just branding: it's an attribute that the agent can use to answer specific queries.
The product schema was always important for SEO. With UCP, it becomes the infrastructure that agents consult throughout the buying process, not just during discovery.
The fields that must be complete and correct:
If the schema has empty fields or partial implementations, it's time to audit and complete it. Not as a future preparation for UCP, but because it affects current performance.
UCP uses the Merchant Center feed as a discovery layer. There are specific attributes that need to be configured:
Google recommends implementing these changes through a supplementary data source in Merchant Center instead of modifying the main feed, to avoid formatting errors that invalidate existing products.
In addition to the standard schema, Google announced, together with the launch of UCP, new attributes in Merchant Center designed specifically for conversational commerce.
These attributes answer questions that agents need to resolve during the selection process:
The logic of these attributes is to reduce model hallucinations during the selection process. When the agent has the information structured and available, they don't have to infer it or search for it in other sources.
The entire technical infrastructure does not work in isolation. AI agents validate product selection by cross-referencing seller data with trusted external sources.
Platforms such as Trustpilot and G2 consistently appear as sources cited by major LLMs.
Positive reviews on these platforms aren't just brand reputation: they're signals that the agent incorporates into the selection process to confirm that the product and the seller are trustworthy.
This doesn't change anything in terms of strategy, but it does reinforce priority. Having an active process of collecting reviews on third-party platforms is part of preparing for UCP, not something separate.
The roadmap published by Google includes expansions that are worth keeping on the radar:
For retail ecommerce teams, the most immediate implications are complex carts and loyalty integration.
For businesses in other verticals, the planned expansion means that the protocol will reach its category sooner rather than later.
The protocol is on the waiting list, but preparation does not require access. The specific steps that can now be taken:
Ecommerce SEO was always about showing up the moment someone is looking for something. UCP extends that principle to an environment where the agent not only shows options, but completes the transaction.
The question worth asking now is whether product data can be selected by an agent, not just appear on a results page.
.png){kind=small}
SEO consulting that scales traffic, grows your brand, increases revenue, and helps you win on Google and AI platforms and LLMs like ChatGPT.
%402x.svg){kind=icon}
