Google Conversational Attributes: Merchant Center Now Feeds AI Mode

On September 16, 2026, Google used its Rethink Retail event and a post by Ashish Gupta, VP and GM of Merchant Shopping, to tell retailers that the Merchant Center feed now shapes how products are found on its AI surfaces. The headline proof point was about Google conversational attributes: during testing with lululemon, attributes the brand submitted were "incorporated 50% of the time in relevant product recommendations in AI Mode."

For a merchant, that changes what a Shopping feed is for. The same Google Merchant Center data that once mostly fed Shopping ads and free listings is now what AI Mode, AI Overviews, the Gemini app and YouTube draw on when a shopper asks a detailed question. A thin feed used to cost you ad efficiency. Now it can cost you the answer.

What did Google announce at Rethink Retail?

Google announced four things retailers can act on now: broader AI performance insights, a YouTube ads Business Agent beta, UCP checkout upgrades, and feed guidance. Here is what each one is, based on Google's own post and its Rethink Retail announcement page.

• AI performance insights, generally available. The AI performance insights report in Merchant Center is now open to businesses in Australia, Canada, India, New Zealand and the US. It compares your share of voice with other brands across AI Mode and AI Overviews and flags attributes missing from your product data.

• Business Agent in YouTube ads. Google is inviting US retailers into a beta that embeds Business Agent inside YouTube ads, so viewers can ask product questions without leaving the video. Business Agent first launched in Search in January 2026.

• UCP integration hub upgrades. Merchants using the Universal Commerce Protocol (UCP) integration hub for checkout can now enable cart transfer to their own site and enhanced checkout flow testing. Google says UCP already supports direct checkout for hundreds of thousands of brands and retailers, with the new capabilities rolling out gradually in the US and Australia and Canada to follow early in 2027.

• Feed guidance. Google pointed retailers to three feed upgrades: the video link attribute for shoppable ad formats, conversational attributes, and loyalty data for member pricing and perks. It also said merchants adopting core feed best practices see a 5% average lift in conversions the following month.

The event page also lists Ask Advisor in Merchant Center, a chat assistant for your feeds, and Checkout Links, which send Demand Gen and YouTube viewers to a merchant's checkout page. Checkout Links are an ads feature and are separate from UCP-powered checkout inside AI Mode and Gemini.

Does your Merchant Center feed really power AI Mode, Gemini and YouTube?

Yes, but this was not a single switch flipped at Rethink Retail; Google has tied more of its surfaces to Merchant Center data throughout 2026. Google's own wording on the event page is that Merchant Center gives you reach on Search, Shopping, YouTube, AI surfaces and more. Pieced together from Google's documentation, the picture looks like this.

Shopify confirms the same routing from the merchant side. Its agentic storefronts documentation says products reach Google AI Mode and Gemini through the Google & YouTube sales channel or Google Merchant Center, not through Shopify Catalog. If you sell on Shopify, your Google feed is the path into Google's AI answers.

What are Google conversational attributes?

Google conversational attributes are six optional Merchant Center fields that give AI systems product context a standard feed lacks, like customer questions, relationships and variants. Google's help page says they complement the main product data specification, help on AI-driven surfaces such as AI Mode and in traditional search, and do not change the approval status of existing products.

Google first announced dozens of new data attributes for AI Mode, Gemini and Business Agent on January 11, 2026, with a small group of retailers. At Google Marketing Live on May 20, it opened conversational attributes to retailers globally. The help center currently documents six.

Google recommends sending these through a supplemental data source, though the primary feed and the Merchant API also work. It also asks you not to repeat what is already in description, product_highlight or product_detail. The value is new context, not restated copy.

What did the lululemon test actually show?

Google says lululemon's submitted conversational attributes were used in half of relevant AI Mode recommendations during testing, which measures usage, not sales or ranking. That is still a meaningful signal. It is the first time Google has put a number on brand-supplied feed context actually appearing inside an AI answer.

Read it carefully, though. The figure tells you that when AI Mode recommended a relevant lululemon product, the extra context lululemon supplied made it into the answer about half the time. It does not say lululemon appeared in 50% of AI answers, that its visibility rose by 50%, or that sales changed. Google has not published the baseline, the number of products or queries, or the test period. Treat it as evidence the fields are read, and measure your own outcome.

Which feed attributes matter most now?

The fields that matter are the ones AI surfaces use to match a detailed question: accurate core data, descriptive attributes, variant structure and conversational context. A shopper asking AI Mode for "breathable running shorts with a phone pocket under $70" is filtering on facts, so every missing fact is a missed match.

Core trust fields

Title, description, link, image, price, availability, brand and unique product identifiers such as GTIN, as defined in the product data specification. If price or stock in the feed disagrees with the landing page, the product can be disapproved, and nothing else you add will matter.

Descriptive attributes

Color, size, material, pattern, gender, age group, product_highlight and product_detail. Google's announcement for AI performance insights names dimensions, weight, materials and colors as the structured attributes it compares against shopper demand, and the report's Popular attributes scorecard lists the ones missing from your data.

Conversational and commerce signals

The six conversational attributes above, plus loyalty data, video links and, if you sell through Google's checkout, the native_commerce attribute. Google's UCP checkout help page says only listings carrying native_commerce(checkout_eligibility) show the Buy button.

How is Google's approach different from OpenAI and Shopify?

Google reads one Merchant Center feed across many surfaces, while OpenAI runs its own product feed spec and Shopify syndicates its catalog to several assistants. For brands that sell everywhere, that means at least two feed formats to keep in sync, which we compare in more depth in our ACP vs UCP explainer.

What should you fix first?

Fix accuracy before adding richness: a disapproved or mismatched product gains little from conversational attributes, so clean core data comes first, then context. Work through this list in order.

1. Clear disapprovals and mismatches. Resolve every Merchant Center diagnostic, and make sure feed price and availability match the landing page.

2. Fill identifiers and categories. Add GTIN, brand and MPN where they exist, and map each product to the right Google product category.

3. Rewrite titles and descriptions for questions. Lead with what the product is, who it is for and the attributes people filter on, not internal names or promo language.

4. Complete descriptive attributes. Color, size, material, pattern, product_highlight and product_detail for every SKU, starting with best sellers.

5. Fix variant structure. Use item_group_id with item_group_title and variant_option so AI Mode sees one product with options, not a pile of near-duplicates.

6. Add question_and_answer from real demand. Pull questions from support tickets, reviews and on-site search. Do not restate the description.

7. Link related products and documents. Use related_product for accessories and companions, and document_link for manuals, size guides and care instructions.

8. Send it all as a supplemental data source. This keeps enrichment separate from your primary feed and easy to update.

9. Check AI performance insights. If you sell in an eligible market, use the Popular attributes scorecard to see which missing fields shoppers care about.

10. Decide on checkout. If you want the Buy button in AI Mode and Gemini, review UCP onboarding in the integration hub and the native_commerce requirement.

For a deeper look at the enrichment side, see our guide to product feed optimization and to structuring product data for AI agents.

What has happened since September?

Since the announcement, Google has reworded its UCP onboarding documentation, while Universal Cart and UK checkout timing remain the open rollout questions for merchants. PPC Land reported on October 7 that Google renamed its help page to "How to onboard to the UCP integration hub in Merchant Center" and removed an FAQ that limited checkout to select US merchants. Google's hub announcement page now describes it as generally available for US selling merchants, while the September post still says gradual rollout.

Universal Cart, announced at Google I/O on May 19, was set to roll out across Search and the Gemini app in the US over the summer, with YouTube and Gmail to follow. For UK brands, Google said in May that UCP checkout would reach the UK after Canada and Australia, and AI performance insights does not yet cover the UK. Conversational attributes, however, are available globally, so UK brands can do the feed work now.

What hasn't Google said yet?

Google has not published how conversational attributes affect ranking, clicks or sales, nor how the lululemon figure was measured or which categories it covers. Several other gaps matter for planning.

• The rest of the "dozens." January's announcement promised dozens of new attributes. The help center documents six.

• Which surfaces read them. The help page names AI Mode "and more." Google has not confirmed how AI Overviews or the Gemini app use conversational attributes.

• Gemini app reporting. Google's announcement page lists the Gemini app as covered by AI performance insights, but the help page describes coverage of AI Mode and AI Overviews only.

• Dates outside the US. No firm dates for UK checkout, or for AI performance insights beyond its five markets.

How Nudge helps brands get ready for AI Mode shopping

Nudge treats the Merchant Center feed as one output of a single enrichment workflow, so the same verified product facts reach Google, Shopify and agents. Here is how that maps to what Google's surfaces now read.

Brands like Mosh, OluKai, and Gorgie are using Nudge to enrich their catalogs & grow agentic revenue by 46% and 8x product mentions.

If you are still setting up or auditing the account itself, start with our Google Merchant Center guide, then come back to the conversational layer. For the bigger shift behind all of this, read what agentic commerce means for brands.