Google Agentic Commerce Update: What Retailers Need to Know

On 16 September 2026 Google announced a set of agentic commerce updates for the holiday season: AI performance insights in Merchant Center are now generally available in five countries, a Business Agent is coming to YouTube ads, the Universal Commerce Protocol checkout tools have new capabilities, and there is fresh guidance on product feeds. Most of the coverage repeated the press release. This guide explains what each update actually does, who can use it, why the UK is missing from most of it, and what a retailer in the UK or the US should change before the peak.

By Moeez Abbas

Published 18 September 2026 · Updated 1 October 2026

Share

• What Google announced, in one table
• What agentic commerce means, and Google’s version of it
• AI performance insights in Merchant Center
• Business Agent inside YouTube ads
• Universal Commerce Protocol: the checkout updates
• The product feed changes every retailer can make now
• What it means for UK retailers
• What it means for US retailers
• How agentic commerce changes SEO and Shopping ads
• How an AI shopping agent decides what to recommend
• The wider agentic commerce market
• The terms in this update, defined
• A 30-day checklist before the holiday peak
• Where this leaves smaller retailers
• Frequently asked questions

What Google announced, in one table

The announcement came from Google’s merchant shopping team on its own blog under the heading “Boost your holiday sales with these agentic commerce updates”. It bundles four things, and they are not equal. Two are new capabilities that change how shoppers interact with your products on Google surfaces; one is a measurement tool; one is a reminder about data quality with a few new attributes attached. The table below separates them, with the availability as Google stated it on the day.

Read that last column again. For the majority of retailers, and for every UK retailer, the actionable part of the “agentic commerce suite” is the feed work at the bottom of the list. The rest is worth understanding because it shows where Google is taking product discovery, but it is not something a Manchester furniture store can switch on this week.

What agentic commerce means, and Google’s version of it

Agentic commerce is shopping in which an AI agent does part of the work a person used to do: finding products, comparing options, checking prices and availability, applying loyalty pricing, filling a basket and, increasingly, completing the purchase. The shopper describes what they want in natural language, the agent researches and recommends, and the transaction happens inside the conversation rather than on the retailer’s website. It is the commercial end of the shift to AI search that has already changed informational queries through AI Overviews and AI Mode.

Google’s version of this has been built in public over 2026. In January it launched the Universal Commerce Protocol, an open standard that lets stores, payment providers and shopping surfaces describe products, prices, carts, checkout, payments and order status in one shared format, so that an AI agent can act across many retailers without a custom integration for each. Business Agent in Search followed, giving brands a conversational assistant that answers shoppers’ questions inside Google’s results. In May, at its marketing event, Google added a Universal Cart that works across Search, the Gemini app and YouTube, finds price drops, and checks out with Google Pay. September’s update extends the same machinery to YouTube ads, adds measurement, and improves the checkout tooling.

The strategic point Google has repeated at every step is that it is not a retailer or a marketplace; it positions itself as the matchmaker between shoppers and brands. In practice that means the retailer still fulfils the order, still owns the customer relationship on paper, and still competes for the recommendation, but the discovery, the comparison and increasingly the checkout happen on Google’s surfaces. For retailers that is both the opportunity and the risk. The opportunity is that a small, well-organised catalogue can be recommended by an agent on merit. The risk is that the agent, not the shopper, reads your product data, and it reads it literally.

AI performance insights in Merchant Center

AI performance insights is the most useful of the four updates for anyone who can access it, because it answers a question retailers have been asking since AI Overviews arrived: are we being shown, and how do we compare? The report sits in Merchant Center and shows how your brand and products are discovered through conversational experiences, comparing your share of voice with other brands across AI Mode and AI Overviews. It was previously in a limited rollout; as of this announcement it is generally available in Australia, Canada, India, New Zealand and the United States. The UK is not on the list, and Google has given no date.

Share of voice here means the proportion of relevant AI answers in which your brand or products appear, set against the competitors that appear instead. That is a different measure from clicks or impressions in Search Console, which count the classic results, and it is the first official measurement of AI visibility for product businesses. Alongside the number, the report is designed to surface actions that improve discoverability, which in practice come back to product data completeness, attribute quality, pricing and availability accuracy, and reviews.

How to use it if you have it

Treat the first month as a baseline, not a verdict. Note the share of voice for your main categories, identify the brands winning the answers you want, and compare their product data with yours: titles, descriptions, images, attributes, price competitiveness, shipping and returns clarity, review volume. Then work through the report’s recommendations in order of the categories that matter most to margin. Re-check monthly rather than daily, because AI answers vary and a single reading tells you little.

What to do if you are in the UK

You cannot see the report, but you can approximate it. A defined set of shopping prompts, checked on a schedule in AI Mode and against AI Overviews, logged with which brands are named and whether yours is among them, gives a manual share of voice that tracks the same thing. It is slower and less precise, and it is exactly what an AI SEO agency does for clients every week. When the official report reaches the UK, the manual baseline will show whether the earlier work moved anything.

Business Agent inside YouTube ads

Business Agent is Google’s conversational assistant for brands. It launched in Search earlier in 2026 so that a shopper who reaches a brand’s result can ask questions and get tailored, instant answers drawn from the brand’s information. The September update invites US retailers into a beta that embeds the same agent inside YouTube ads. A viewer watching an ad can research a product, ask complex questions and receive guidance without leaving YouTube, moving, in Google’s words, from inspired to purchase-ready inside the video context.

For advertisers this changes what a video ad is. It stops being a message and becomes the front door to a conversation, and the quality of that conversation depends on the information the agent can draw on: product data, policies, sizing, compatibility, delivery times, returns, stock. An ad that sends a curious viewer into an agent that cannot answer “does this fit a 2019 model” or “will it arrive before the 24th” will lose the sale in the ad itself. The preparation is the same as for every other part of this update: complete, accurate, structured product information, plus clear answers to the questions your customer service team already hears most.

Availability is narrow for now: the US, retailers who sign up, and only within eligible campaigns. UK advertisers cannot join yet. What they can do is watch the results US retailers report, and make sure their own product data and FAQ content would survive an agent reading it aloud to a customer.

Universal Commerce Protocol: the checkout updates

The Universal Commerce Protocol, UCP, is the foundation of Google’s agentic commerce. It already powers direct checkout for hundreds of thousands of brands and retailers across Google, letting a shopper buy inside Search, including AI Mode, and inside the Gemini app, with the merchant fulfilling the order. Merchants connect through the UCP integration hub in Merchant Center. The September update adds two capabilities to that hub for merchants using UCP-powered checkout: cart transfer to the merchant’s own site, so a basket built on Google can be completed on the retailer’s checkout when that is the better route, and enhanced checkout flow testing, so merchants can verify the buying journey before customers hit it during peak. Analytics for these flows are promised next.

The rollout is US-first, with Australia and Canada to follow early next year. UK retailers are not included, and the wider UCP checkout is not something a UK Merchant Center account can enable today. That does not make the protocol irrelevant to UK businesses; it makes it a roadmap. When agentic checkout reaches the UK, the merchants that will be ready are the ones whose feeds are already complete, whose stock and pricing are already accurate in near real time, and whose returns and delivery policies are already machine-readable.

What cart transfer and checkout testing solve

Cart transfer addresses the retailer’s biggest objection to agentic checkout: losing the customer at the moment of purchase. If the basket can move to the merchant’s site, the retailer keeps the payment relationship, the upsell and the account creation where it wants them, while still being discoverable through the agent. Checkout flow testing addresses the operational fear: a checkout that breaks inside an AI surface during the last week of November is invisible to your own monitoring. Both are the kind of unglamorous plumbing that decides whether agentic commerce is a channel or a leak.

The product feed changes every retailer can make now

The part of the announcement that applies everywhere, immediately, is also the part most retailers skip. Google’s own figure is that merchants adopting the core Merchant Center feed best practices see, on average, a 5% increase in conversions the following month. That is a Google statistic about Google’s advertisers, so treat it as a direction rather than a promise, but the direction is consistent with every audit of a Shopping account: the feed is where the money is lost. Four items are specifically called out.

Core feed best practices

Accurate titles that lead with what the product is, then brand, then the attributes people search by; descriptions that describe rather than sell; every applicable attribute filled, from colour, size and material to GTIN, condition and age group; images that meet the guidelines and show the product on a plain background with lifestyle variants added; price and availability updated as often as they change, ideally through the API rather than a daily file; shipping and returns policies set at account level and overridden per product where they differ. None of this is new, and almost no feed passes all of it.

The video link attribute

Submitting a video link for a product lets Google power visual, shoppable ad formats automatically from the feed. For retailers with product videos already on their site, this is a matter of adding a column; for those without, short product demonstrations shot on a phone are enough to start. Video matters more as agents and AI surfaces look for richer context to recommend a product with confidence.

Conversational attributes

Conversational attributes let a merchant add context that answers the questions a shopper asks an agent: who the product suits, what it pairs with, what problem it solves, what it is not for. In Google’s testing with a large apparel brand, conversational attributes submitted by the brand were incorporated in relevant AI Mode product recommendations about half the time. That is a striking number, because it means the retailer can feed the agent the language it uses to recommend the product, rather than hoping the agent infers it from a title. Writing these attributes well is copywriting for machines: specific, honest, in the words customers use.

Loyalty data in Merchant Center

Connecting a loyalty programme to Merchant Center lets Google show member pricing and perks, such as free shipping, in results for signed-in members, drive repeat visits and steer bids towards the highest-value shoppers. For retailers with a loyalty scheme it is a straightforward way to make agentic and conventional results reflect the real price a member pays, which is often the price that wins the comparison.

What it means for UK retailers

The honest summary for a UK retailer is that three of the four updates are not available to you, and the fourth is where your advantage lies. AI performance insights, Business Agent for YouTube ads and the UCP checkout capabilities are all launched in the US, with Australia, Canada, India and New Zealand ahead of the UK on the insights report. Google has not given a UK date for any of them. AI Mode and AI Overviews are live in the UK and already shape product discovery, so the visibility question is real now; the official measurement and the agentic checkout are not.

That leaves a clear plan. Fix the feed to the standard above, because it improves Shopping ads, free listings, AI Overviews and AI Mode today, and it is the prerequisite for everything that arrives later. Add conversational attributes to the products that matter most, because AI Mode recommendations in the UK draw on the same data. Build the manual share-of-voice tracking so you know where you stand in AI answers before the official report exists. Make product pages and policies machine-readable with structured data, since agents and AI surfaces read schema markup for price, availability, ratings and returns. And keep the conventional work honest: the map pack and organic listings still take most UK retail clicks, and a strong SEO foundation is what the AI surfaces retrieve from.

What it means for US retailers

US retailers have the full set, which makes the sequence matter more than the features. Start with the feed, because the insights report will otherwise measure a catalogue that is not ready, and the YouTube agent will answer from incomplete data. Then open AI performance insights, record the share of voice by category, and pick the two or three categories where a gain is worth the most margin. If YouTube is a real channel for you, apply to the Business Agent beta and prepare the information it will need. If you already run UCP-powered checkout, use the new flow testing before the peak and decide, category by category, whether cart transfer to your own site or checkout on Google converts better; the analytics that Google says are coming will make that a data decision rather than a guess.

Budget for the fact that agentic surfaces compete with your own site for the transaction. A shopper who buys inside the Gemini app or AI Mode is a customer you may never see on your site, so the loyalty data connection and the account creation on cart transfer are the levers that keep the relationship. And measure the whole thing on booked orders and margin, not on impressions in an AI surface; the point of the update is sales during the gifting season, and the report that matters is the one that ties AI visibility to revenue.

How agentic commerce changes SEO and Shopping ads

The pattern across all four updates is that product data has become the content. In classic SEO the page ranked; in agentic commerce the product record, the reviews and the policies are what the agent reads, compares and recommends, and the page is where it sends the customer, if it sends them anywhere. That changes priorities for both organic and paid work.

On the organic side, product pages need structured data that agrees with the feed: Product schema with price, availability, brand, GTIN and aggregate ratings; Organization schema with the policies; FAQ content that answers the questions the agent will be asked. Category and buying-guide content earns the citation in AI Mode when the shopper is still deciding what to buy, which is where a specialist retailer beats a marketplace. Reviews on the platforms Google reads decide whether the agent trusts the product. And the technical basics, crawlable, fast, one canonical version of each product, decide whether the data is retrieved at all.

On the paid side, Shopping and Performance Max campaigns run on the same feed the agents read, so every feed fix improves both. Conversational attributes and video links give the automated campaigns more to work with. Loyalty pricing changes which shoppers see which price. And the arrival of agents inside ads means the ad’s job is to start a conversation the data can finish. The Google Ads management here already treats the feed as the campaign, and the September update makes that the only sensible way to run it.

How an AI shopping agent decides what to recommend

Every update in this announcement makes more sense once you picture what the agent is doing. A shopper asks for “a waterproof jacket for a Lake District walking holiday under £150 that will arrive by Friday”. The agent breaks that into constraints: category, use case, price ceiling, delivery date, and probably size and colour once it asks. It then retrieves candidate products from the surfaces it can read, which for Google means Merchant Center product data first, product pages and reviews second. It scores them on how completely and confidently they satisfy the constraints, and it recommends two or three with reasons. If the shopper says yes, it either hands them to the retailer or, where checkout is enabled, completes the purchase.

Three things follow from that sequence. First, a product with a missing attribute is not marked down; it is excluded, because the agent cannot confirm it meets the constraint. Second, real-time pricing and availability are not a nicety; a product that shows as in stock and is not will be recommended, fail at checkout, and cost you the trust of both the shopper and the agent. Third, the reasons the agent gives are drawn from your data and from what other people say about you, which is why conversational attributes and product reviews carry so much weight. The retailer that writes “suits beginners, pairs with our base layer, machine washable, 30-day returns” in the words a customer would use has handed the agent its recommendation script.

The wider agentic commerce market, briefly

Google is not the only company building this, and the retailer’s job is to be legible to all of them without rebuilding for each. Payment networks, checkout providers, ecommerce platforms and the other AI assistants have all published their own agent protocols and shopping integrations over the past year; some have been launched, paused and relaunched. Google’s answer to the fragmentation is to make the Universal Commerce Protocol open, so that the same product, cart and order formats work across surfaces, and to pair it with an agent payments protocol for the payment step, so that an agent can pay with the shopper’s authority and the merchant can trust that it did. Whether one standard wins or several coexist, the merchant-side requirements are identical: structured product data, live pricing and stock, machine-readable policies, review data and a checkout that works when a machine drives it.

Two adjacent shifts deserve a line each. Retail media, the advertising sold by retailers on their own sites, is being pulled into agentic flows, and the platforms that own the agent will decide what a sponsored recommendation looks like; expect Google’s Shopping ads and Performance Max to be the first place that shows up. And first-party data, the loyalty programme, the purchase history, the customer service transcripts, becomes the retailer’s main defence, because it is the one thing the agent cannot get from a competitor’s feed. Connecting loyalty data to Merchant Center is the first visible example of that data earning its keep inside an agentic surface.

The terms in this update, defined

Agentic commerce is shopping in which an AI agent researches, compares, recommends and, increasingly, buys on a person’s behalf. Agentic AI is the broader category of AI systems that take actions towards a goal rather than only answering. An AI shopping agent is the assistant that does this for products; Google’s surfaces for it are AI Mode, the conversational search experience, AI Overviews, the generated summaries at the top of results, and the Gemini app. The Universal Commerce Protocol, UCP, is Google’s open standard for describing products, carts, checkout, payments and orders so agents and merchants can transact; the UCP integration hub is where a merchant connects to it in Merchant Center, Google’s product data platform. Universal Cart is Google’s cross-surface basket with Google Pay checkout. Business Agent is the brand-specific conversational assistant Google offers in Search and now in YouTube ads. AI performance insights is the Merchant Center report measuring how products are discovered in AI surfaces, expressed as share of voice against other brands. Conversational attributes are natural-language product details submitted in the feed for AI recommendations; the video link attribute supplies product video for shoppable formats; loyalty data connects a membership programme so member pricing and perks appear. Generative engine optimisation, GEO, is the practice of making a brand and its products visible in AI-generated answers, which for retailers now runs through product data as much as through content.

A 30-day checklist before the holiday peak

This is the order the work is done here for retail clients, whatever the country, between now and the November peak.

Week one: the feed

Run the Merchant Center diagnostics and clear every disapproval and warning. Rewrite the titles and descriptions of the top hundred products by revenue to the best-practice pattern. Fill every missing attribute for those products. Confirm price and availability update automatically, and test a price change end to end. Set shipping and returns at account level and check the exceptions.

Week two: the machine-readable layer

Validate Product, Offer, AggregateRating and Organization schema on the top products and the policy pages, and make it match the feed exactly. Add conversational attributes to the top products in the words customers use. Add video links where videos exist. Connect loyalty data if a programme exists. Submit product reviews through the Merchant API if reviews live in a third-party system.

Week three: visibility and answers

Baseline AI visibility: the official report where available, a manual prompt set everywhere else. Write or update the buying guides and FAQs for the categories that matter, with direct answers an agent can quote. Check the stock and delivery-cutoff messaging the agent will repeat to shoppers in December.

Week four: paid and testing

Refresh Shopping and Performance Max assets from the improved feed. US merchants: run the UCP checkout flow tests and apply to the YouTube Business Agent beta if relevant. Everyone: set the weekly report to show orders and margin by surface, so the peak is judged on money rather than mentions.

Where this leaves smaller retailers

Announcements like this are written for large advertisers, and the case studies Google cites are national brands with data teams. The reassuring part for a smaller retailer is that the machinery rewards data quality more than budget. An agent recommending a product does not know your ad spend; it knows your title, your attributes, your price, your stock, your reviews and your policies, and a ten-person business can get those right faster than a national chain with a legacy catalogue. The less reassuring part is that the features that let you compete on the agentic surfaces themselves are arriving country by country, and the UK is not first. The right response is neither to ignore the update nor to wait for it: do the feed and structured data work now, because it pays in the channels you already have, and be ready when the rest arrives.

It is also worth being clear about what this update does not do. It does not change how the map pack or organic results work for a local retailer with a shop, it does not replace the product page, and it does not mean AI agents are about to buy on behalf of most shoppers this Christmas. Agentic checkout is real and growing, but for the peak of 2026 the majority of purchases will still go through search results, Shopping ads and the retailer’s own site, all of which the same feed work improves.

Frequently asked questions

What did Google announce in the September 2026 agentic commerce update?

Four things: AI performance insights in Merchant Center are generally available in Australia, Canada, India, New Zealand and the US; a Business Agent for YouTube ads beta in the US; new capabilities in the Universal Commerce Protocol integration hub, cart transfer and checkout flow testing, rolling out in the US with Australia and Canada early next year; and product feed guidance covering best practices, the video link attribute, conversational attributes and loyalty data.

Is Google’s agentic commerce update available in the UK?

Mostly not. The insights report, the YouTube agent and the UCP checkout capabilities were announced for other countries and Google gave no UK date. The feed guidance applies to every Merchant Center account, including UK ones, and AI Mode and AI Overviews are already live in the UK.

What is agentic commerce?

Shopping in which an AI agent finds, compares, recommends and increasingly buys products on a person’s behalf, based on a natural-language request. Google’s implementation runs on the Universal Commerce Protocol, with checkout inside Search, AI Mode and the Gemini app, and a Business Agent that answers shoppers’ questions for brands.

What is the Universal Commerce Protocol?

An open standard Google launched in January 2026 that describes products, prices, carts, checkout, payments and order status in one shared format, so AI agents and shopping surfaces can transact with many retailers without custom integrations. Merchants connect through the UCP integration hub in Merchant Center.

What are AI performance insights in Merchant Center?

A report that shows how your brand and products are discovered through AI Mode and AI Overviews, including share of voice against other brands, with recommendations to improve discoverability. It is generally available in five countries as of September 2026.

What are conversational attributes in a product feed?

Extra context a merchant submits about a product, such as who it suits, what it pairs with and what problem it solves, written in natural language so AI surfaces can use it in recommendations. In Google’s testing with one apparel brand they were incorporated in relevant AI Mode recommendations about half the time.

Does agentic commerce mean shoppers stop visiting my website?

Some will buy inside Google’s surfaces. Cart transfer lets a basket built on Google complete on your site, and loyalty data keeps member pricing visible, which are the levers for keeping the relationship. For the 2026 peak most purchases will still arrive through search results, Shopping ads and your own site.

What should a retailer do first?

Fix the product feed: disapprovals, titles, attributes, images, price and availability sync, shipping and returns. Google’s own figure is an average 5% conversion lift the following month from the core best practices, and every other update depends on the same data.

How does this affect Shopping ads and Performance Max?

They run on the same feed the agents read, so feed improvements lift both. Conversational attributes, video links and loyalty pricing give automated campaigns more to work with, and the Business Agent in YouTube ads turns a video ad into a conversation the product data has to be able to finish.

How do I measure AI visibility if the insights report is not available in my country?

Define a set of shopping prompts for your categories, check them on a schedule in AI Mode and against AI Overviews, log which brands are named and whether yours is, and report the trend monthly. It is a manual share of voice and it is close enough to act on.

Want your feed and AI visibility checked before the peak?

Send your website and your Merchant Center country, and you will get a written review: the feed problems that are costing conversions, whether your product data would survive an AI agent reading it, where your brand appears in AI Mode and AI Overviews for your categories, and what to fix in which order before November. No jargon, no lock-in, no passwords.

Send the details through the contact form, or message me on WhatsApp at +44 7853 133869.