AI search is becoming an operating surface for brand teams. The practical question is not whether a single model response changed—it is whether the change reveals something your team can verify and improve.
What Google has made available
Google’s September 16 holiday-commerce update highlighted that AI performance insights in Merchant Center are now available to eligible accounts in Australia, Canada, India, New Zealand, and the United States for English-language queries. The report is designed for shopping-related conversational discovery in AI Mode and AI Overviews. It gives merchants a first-party view of how their products and brands appear in those surfaces, with data organized around shopping stages, terms, attributes, and intent. This is not a new ranking system or a promise of a particular result. It is a new place to inspect Google’s own record of eligible organic AI-shopping exposure.
The report measures a narrower surface than a visibility claim
Google says the report covers organic AI traffic, such as free listings, and excludes paid Ads traffic. That boundary matters. A product’s presence in the report is not a count of all Google shopping impressions, and it should not be combined automatically with Shopping campaign data, paid placements, or traffic from other answer engines. It also does not show a ChatGPT or Perplexity result. Treat it as a platform-specific observation of how products appeared in supported Google AI shopping experiences, under the report’s own definitions.
What merchants can actually inspect
The report groups conversational shopping activity into discovery, evaluation, and ready-to-buy stages. It also exposes top terms, popular attributes, search intents, product counts, and a share-of-voice measure against the competitors Google defines for the account. In Google’s documentation, share of voice is the brand’s AI impressions divided by the total AI impressions of that brand and its defined competitors for related queries. That is useful for comparing movement inside the selected Merchant Center cohort. It is not a universal market-share figure, an answer-quality score, or proof that buyers preferred the brand.
Product data has become more visible in the measurement loop
The accompanying Google update emphasizes product-feed completeness: conventional attributes, video links for shoppable formats, and conversational attributes that add product context. Google reports that, in testing with lululemon, merchant-submitted conversational attributes were incorporated in 50% of relevant AI Mode product recommendations. That is a platform-reported test result, not a general performance benchmark and not evidence that adding a field will produce the same outcome for another merchant. It does, however, make the input-output chain more inspectable: missing or weak product data can now be reviewed beside observed AI-shopping exposure.
Do not turn a metric into a recommendation claim
An AI impression records an appearance, not the language a shopper saw around it. A product can be surfaced without being the first choice; it can appear in a comparison without receiving a positive recommendation; and it can be available for a conversational question without attracting a click or transaction. Google’s Merchant Center report does not expose the full response text, every source used by an AI answer, or a shopper’s reasoning after the interaction. Teams should keep answer captures, customer research, referral behavior, and conversion records alongside the report rather than treating one dashboard field as the complete discovery story.
Use the time and competitor limits carefully
Google says historical data is updated daily with a lag of a few days. It also notes that accounts without enough competitor data can show a 100% share of voice, and that an account’s share and its competitors’ average share can both rise when the underlying competitor data changes. Those details make the report more useful, not less—but they mean a one-day swing should not trigger a feed rewrite or an executive claim. Establish a baseline by category, preserve the reporting date and market, and investigate repeated movement before assigning a cause.
The practical response is feed hygiene plus a separate evidence record
For commerce teams, start with the facts a buyer needs to make a decision: accurate titles, availability, price, variant details, material, dimensions, compatibility, delivery and returns information, and relevant use-case context. Use the report’s missing-attribute and intent views to prioritize gaps, then document the exact product-data change and when it became available. Separately, monitor the visible answer or shopping experience for important buyer questions, including which products appear, the qualifiers attached to them, and any route to the merchant. This keeps a manageable test from becoming an unsupported causal story.
What remains unknown
Google has not published a complete list of every prompt, ranking input, competitor-set rule, or response layout used in these AI-shopping experiences. The report is currently limited to eligible Merchant Center accounts and English-language queries in five countries. It does not establish how a product will appear in every AI Mode or AI Overview response, whether a particular feed edit caused movement, or whether an appearance produces incremental revenue. The 50% conversational-attribute figure is likewise a Google test observation, not a cross-merchant guarantee.
Bottom line
Google’s Merchant Center AI performance insights give eligible retailers a more concrete way to inspect organic visibility in AI shopping across AI Mode and AI Overviews. Their value is specificity: product data, shopping stage, intent, and a platform-defined comparison set can now be examined together. Use that evidence to improve verified product information and run controlled checks. Keep paid performance, answer framing, recommendations, and commercial outcomes in their own records.