Does Merchant Center data power Google's AI shopping?

Google maintains an understanding of products that spans retailers, manufacturers and marketplaces, and merchant feeds are one of its most direct inputs. When a shopping answer names options, prices and availability, it is drawing on that understanding rather than reading one retailer's page in isolation.

The practical implication is that eligibility for those surfaces starts in the account rather than on the website. A product with no feed entry can still be understood from crawled pages and structured data, but the retailer has given up control over how its price, stock, shipping and attributes are represented.

Free product listings are the specific reason to care. They are the unpaid presence in Shopping and related surfaces, they are configured in Merchant Center, and they require the same feed quality standards as paid campaigns. A store that runs ads but has never checked whether free listings are enabled is leaving a channel switched off.

Account health governs all of it. A suspension, a policy issue or a large group of disapproved items removes products from consideration on every surface at once, which is why the diagnostics view is the first thing to check when a catalogue disappears from anywhere in Google's shopping experience.

Feed completeness stops being a Shopping detail and becomes the basis of visibility. Identifiers first: GTIN where one exists, brand, MPN. Then the attributes that let a product be matched to a specific request: category, product type, colour, size, material, gender, age group, condition, and the group identifier tying variants together.

Price, availability and shipping must agree with what the landing page shows when Google fetches it. Google's automatic item updates can correct some mismatches from structured data on the page, which is a safety net rather than a substitute for a feed that is already right.

Shipping and returns settings deserve specific attention because shoppers increasingly ask about them in the question itself. Configure delivery times and returns policies in the account, and keep them consistent with the policy pages on the site.

Image quality is the other under-managed input. Product panels are visual, and the primary image needs to be clean, well lit and free of added text or promotional overlays. Additional images give context.

Refresh frequency matters more than it used to. A daily fetch is adequate for a stable catalogue; a store with fast-moving stock or frequent price changes should be pushing updates continuously through the content API or a real-time integration.

ChatGPT, Perplexity, Claude and Copilot have no access to your Merchant Center data. They build their picture from pages their fetchers can retrieve, from search infrastructure they license or operate, from marketplaces and from whatever merchant integrations they have established directly.

Microsoft is the closest parallel, since it operates its own merchant program that supplies structured product data for Bing shopping experiences and the Copilot surfaces built on them. A store selling in Canada or the United States that already maintains a Google feed usually finds that submitting the same data there is a modest amount of extra work.

Several assistant providers have introduced their own routes for merchants to supply product information, and those arrangements change frequently enough that the only reliable answer is to check each provider's current merchant documentation rather than to trust a description written months ago.

What carries across all of them is the underlying data hygiene. Correct identifiers, attributes that exist as fields rather than as sentences, prices and stock that match the page, and consistency between your site, your feeds and the marketplaces that list you. A store that gets that right is a good candidate everywhere, without needing to guess at any single system's mechanics.

Treat it as a data quality program with a fixed checklist and a schedule, not as a new channel. Once a month, work through the diagnostics view and clear disapprovals by type rather than by item, because the same missing attribute usually explains hundreds of them at once.

Once a quarter, audit attribute coverage: what share of the catalogue has a GTIN, a populated size and colour, a product type, a group identifier, more than one image. Coverage gaps are the most common reason a product is absent from constraint-heavy answers, since a request that names an attribute you have not supplied cannot match your item.

Spot-check the round trip. Take ten products, look at the feed values, open the landing page, and confirm price, availability and shipping agree. Do the same after any theme change, app installation or pricing rule.

Then measure what can be measured: free listing clicks and impressions in Merchant Center performance, Shopping campaign data for the paid side, and site revenue. Presence inside a specific AI answer is not reportable, so use those numbers to judge the account and treat the answer surfaces as a consequence of doing this well.