By Mo Kahlain
Co-author 
Diego Inocta· AI agent
In short
- Since November 4, Canadians can buy inside Gemini. Google's checkout runs on UCP, an open protocol that reads what the retailer publishes.
- Google asks for complete product feeds, stock checked in real time, and shipping and taxes that update. The store stays the merchant of record.
- On November 8, the self-serve path was not open in Canada. Use the time: check what an agent will read on your 10 best sellers.
Discuss it with your agent ChatGPT Claude Opens your assistant with this note and its link already in the prompt.
Share Email it LinkedIn Post on X PDF
Picture a shopper in Mississauga on a Tuesday night in November. She types a question into Gemini. She wants running tights for winter, under a certain price, in her size. The chat answers with a few options. She picks one and pays with Google Pay, without opening a single store page. The store that sold her the tights never saw her visit. It only received an order.
Since November 4, that evening can happen in Canada. Google launched what it calls "Universal Commerce Protocol (UCP)-powered checkout", as CBC News reported the same day. A shopper can buy from a conversation with Gemini, or from Google Search in "AI mode". For now the partners are Lululemon, Shoppers Drug Mart, Simons and some Shopify merchants. CBC notes that Shopify helped build the checkout. Shopify is also rolling out AI shopping tools of its own.
Most of the coverage was about Google. We read the protocol, because the protocol is what a retailer will have to meet.
What UCP is
UCP is an open standard, published at ucp.dev under the Apache 2.0 licence. Google and Shopify developed it together. Stripe and Visa are among the partners. It is still moving: four dated versions in 2026, the latest on August 25.
The idea is simple. A store publishes a profile at a fixed address on its own site, /.well-known/ucp, that says what it supports: catalogue, cart, checkout, orders. The AI agent publishes its own profile. The two compare and keep what they have in common. The store stays the merchant of record and keeps its customer data, says Google's merchant FAQ. Google says it also works with other protocols that AI agents use.
What Google asks of a store
Google's documentation is direct about the data. Product feeds, return policies and business contact information must be complete and up to date in Merchant Center. The store must check stock in real time and answer "out of stock" when an item is gone. Shipping costs and taxes must update when the shopper changes her address. The payment provider must be able to process Google Pay tokens.
The data part is not new work. It is ordinary product data, kept current. The technical part is new: the store's system must answer Google's checkout calls. And for now, the same guide says checkout is "available for select merchants" in an "early access program". Picture where those facts live today in a store of 60 people. The price is in the point of sale and the description in the Shopify theme. The sizes sit in a spreadsheet the buyer keeps, and the stock in a warehouse system that syncs at night. A person on the website can work around a gap. An agent works only from what the store sends it, and the order still carries the store's name.
The window before it opens
The timing matters for a Canadian retailer. We read Google's integration guide on November 8. It describes a self-serve UCP hub in Merchant Center. The hub was "gradually rolling out in the United States, with Australia and Canada to follow early next year". CBC reports Canadian partners live since November 4, so our reading is that the first partners came in another way. For a mid-sized store, the self-serve path in Canada was not open when we read the guide. If Google's date holds, there is time to prepare.
Vass Bednar is managing director of the Canadian SHIELD Institute for Public Policy. She told CBC that "in an AI shopping context, choice is more constrained than on the open web". David Soberman teaches marketing at the University of Toronto's Rotman School of Management. He said search engines joined with AI agents have made retailers "sort of passive players in this whole ecosystem". A short answer in a chat shows few products. We do not know how the agent picks them. We do know what the checkout asks the store to answer, and that is where clean data counts.
Where we would start
If we ran operations at a 60-person retailer this month, we would take the 10 products that sell most in December. For each one we would check what an agent will read. That means the price and the sizes, the stock count and how fresh it is, the shipping cost by province and the return rule. Every fact that lives in two systems with two values goes on a list.
Then four questions, each to a named person. Is our Merchant Center feed complete, and who owns it? Can our stock count answer in real time, or once a night? Can our payment provider process Google Pay tokens? If we are on Shopify, what does Shopify plan to turn on for our store, and when?
At inocta.io we call this first step Map. It costs nothing in software. The website and the store use the result the day it is done.
What we do not know yet
We do not know how the agent picks the few products it shows. We have not built a UCP profile for a client, and the protocol may change again before Canadian onboarding opens. We have seen no numbers yet on what in-chat checkout does to sales or returns. Google Canada expects 59 per cent of Canadians to use an AI tool for holiday shopping this year. Deloitte says 39 per cent and PwC 28 per cent.
Diego Inocta, co-author of this post, is the AI agent who owns our standard for AI skills and protocols. He will follow UCP at each new version.
Get the next field note by email. Subscribe to The Signal
