Run product data like a weekly release, not a one time feed cleanup

What should an ecommerce team do with AI shopping signals?

Use AI shopping signals to find a product fact that needs work, then update the record that owns the fact before you change marketing copy. The useful unit of work is a category release, not a vague request to "optimize for AI."

Google's current Merchant Center documentation describes AI performance insights as a limited United States pilot. The report focuses on organic AI traffic and covers share of voice, shopping journey phases, product terms, and structured product attributes. That scope makes it a diagnostic tool. It cannot show every customer journey, paid performance, or downstream revenue result.

Start with a pattern your team can investigate. A popular feature term with low visibility may point to a missing specification. A repeated question in support may show that the product page explains the product poorly. A high return rate may reveal an important fit condition that sits only in an internal document. Find the fact, name its owner, and repair it in the source systems that publish it.

Where should the product facts match?

Customers do not experience your catalog as a feed. They encounter a product page, a shopping result, a review, a delivery question, and an order confirmation. Your data process should check the places where a material fact can drift.

Google's Merchant Center guidance says structured data on a landing page can help retrieve current product information from your site. It also warns that the markup and product data need to match. Treat that match as a release check, not a cleanup after an item is already live.

When do conversational attributes earn their place?

Conversational attributes can help an ecommerce team add answer shaped product facts, but they are optional. Google's documentation lists question and answer, document link, related product, item group title, variant option, and popularity rank. It also tells teams not to duplicate facts already supplied in descriptions, product highlights, or product details.

Use them where customers need a precise answer. A compatible accessory, a size constraint, a material tradeoff, or a setup requirement can change a purchase decision. Do not fill these fields with keyword lists, vague sales language, or the same claim repeated in five locations. Each field should settle a real question and lead back to a product fact your business can prove.

How can a retailer test one category without creating a large project?

Choose a category with enough volume to show real customer behavior and enough complexity to reveal a useful problem. Items with variants, compatibility rules, technical details, or frequent support contacts make good candidates. Avoid a category in the middle of a major platform migration.

List the ten questions customers ask before purchase. Mark each one as answered, partly answered, unsupported, or contradicted by another public source. Then check a small sample of product pages, the feed, product structured data, shipping and return policies, and support articles. Fix the product record first. Publish the aligned change in the same release.

For sites using merchant listing markup, July 7, 2026 added the category property to the documented attributes. That does not turn a category into a magic AI signal. It does give teams another reason to keep category language consistent between the merchant feed, page architecture, and visible product context.

Which evidence can support AI shopping discovery beyond your own site?

A product page is only one public record. For many product categories, buyers and AI systems can encounter manufacturer documentation, authorized retailer listings, product ratings, support manuals, structured return information, and authentic customer reviews. Those sources should reinforce the same core facts without copying one another.

That is the citation environment for ecommerce. A business should not try to manufacture it with inflated claims or copied reviews. Keep product identity, feature language, availability, and policy terms accurate wherever your company controls them. Earn independent proof through reliable delivery, useful support, and a legitimate review process. When outside records conflict with your site, find the source of the mismatch instead of publishing another explanation.

What should leaders measure after the release?

Measure changes your team can act on. Track product data issues, support contacts about the selected questions, return reasons, page corrections, feed warnings, and conversion quality for the category. If the Merchant Center pilot is available to your account, record its organic visibility signals beside those operating measures. Do not treat a share of voice number as revenue attribution.

Google's report documentation gives a second caution: a zero share of voice can mean there were not enough impressions, and a 100 percent value can occur when the account lacks defined competitors. Read the number with the report's limits in mind. A better outcome is a catalog that gives a customer fewer surprises, even if no dashboard produces a dramatic graph.

Frequently asked questions about AI shopping product data

What are AI performance insights in Merchant Center?

They are a Merchant Center pilot for visibility in specific Google generative AI shopping experiences. The report covers organic AI traffic, not paid traffic or complete commerce attribution.

Should ecommerce teams add conversational attributes to every product?

No. Start where buyer questions or technical details create friction. Keep each answer accurate, useful, and distinct from information already present in the product record.

Does accurate product data guarantee AI shopping visibility?

No. Accurate product data helps your team publish consistent facts, but it cannot guarantee an AI result, a recommendation, traffic, orders, or revenue.

Sources

• Google Merchant Center Help, About AI performance insights. Used for availability, report scope, metrics, filters, and data limitations.
• Google Merchant Center Help, How to use conversational attributes. Used for the optional attributes and duplication guidance.
• Google Merchant Center Help, Supported structured data attributes and values. Used for alignment between page markup and Merchant Center product data.
• Google Merchant Center Help, Tips to help your products stay approved. Used for landing page and product data consistency.
• Google Search Central documentation updates. Used for the July 7, 2026 merchant listing category update.
• Google Search Central, optimizing for generative AI features. Used for the broader context that search fundamentals remain relevant for generative AI features.