Merchant Center’s New AI Performance Insights Report, Explained

Google Merchant Center’s AI performance insights report is the first time Google has given merchants a native window into how their brands and products actually surface inside AI Mode, AI Overviews, and the Gemini app. The pilot went live for a limited set of US accounts around July 13–14, 2026, and it answers a question ecommerce teams have been guessing at for over a year: am I visible in conversational shopping results at all?

The stakes are simple. AI Mode alone passed one billion monthly users within a year of launch, per Google — yet until this pilot, merchants had no first-party data on whether their catalog appears in those conversations. Third-party visibility tools filled the gap with estimates. This report replaces estimates with Google’s own impression data, at least for the accounts that have it.

This guide covers what shipped and where to find it, the two-stage timeline from Google Marketing Live to pilot, the four documented metrics and their exact definitions, the scope limits Google states outright, the gaps it doesn’t address — clicks above all — and a practical preparation checklist for merchants still waiting on access. The angle throughout is measurement, not another AI Overviews optimization lecture.

The feature is called AI performance insights, and it lives in a new tab: Merchant Center → Analytics → Products → AI performance. It is visible only to accounts included in the pilot — there is no documented application process. Google’s help documentation describes the purpose as helping merchants “understand how your brands show up in conversational results, evaluate your visibility across different phases of the shopping journey, and view trends to help optimize your product data.”

The report was first spotted by Hana Kobzova of PPC News Feed on July 13, with Search Engine Roundtable’s coverage and Google’s help document, “About AI performance insights,” following on July 14. SEO consultant Brodie Clark shared screenshots of the live dashboard, which Search Engine Roundtable referenced as confirmation the pilot is functioning in at least one account.

Scope, from Google’s own help doc: “AI performance insights is currently in a pilot with a limited number of Merchant Center accounts in the US. This feature will roll out more broadly, expanding to Australia, Canada, India and New Zealand in the coming months.” No firm date accompanies that expansion — worth keeping in mind before anyone puts it on a client roadmap. The pilot also lands in a busy season for the platform: Merchant Center has been absorbing naming and platform changes through 2026, and this report is the most substantive addition to its analytics surface so far.

Most coverage treats this as a single July event. It isn’t. The AI Performance Insights concept — alongside a companion feature called Conversational Attributes — was announced at Google Marketing Live 2026 on May 20, per Search Engine Land. The hands-on pilot appeared in accounts roughly eight weeks later, on July 13–14. That gap matters for planning: Google previewed the measurement story two months before any merchant could touch it, and the four-country expansion is still running on the same “coming months” language today.

The two features split cleanly along a read/write line. Conversational Attributes is the input: retailers add conversational product attributes and richer descriptions directly in Merchant Center, which Google’s AI systems use to match products to conversational shopping queries across AI Mode, Gemini, and other AI surfaces — and it is rolling out globally. AI performance insights is the output: the report showing whether that product data is actually earning visibility. The asymmetry is notable — the write side is global, the read side is one country.

Semrush’s coverage frames the report as the first time Google has offered native reporting on AI search visibility metrics for ecommerce — a reasonable framing, though it comes from a vendor that sells competing visibility tooling. Semrush also claims Bing shipped its own AI performance report in Webmaster Tools earlier in 2026; we found no second source for that claim in our research, so treat it as Semrush’s assertion rather than established fact.

Google’s help documentation defines four metrics and dimensions for the report. All of them are built on the same underlying unit — AI impressions — which is why the report reads as a visibility dashboard rather than a performance dashboard. The Share of Voice definition is the one worth memorizing, because it’s the number executives will ask about: “The % is based on your number of AI impressions divided by the total amount of impressions across your competitors and yours shown for related queries.”

Around those four metrics, Google’s pre-announcement documentation organizes insights along three additional axes: shopping journey stages (discovery, evaluation, purchase), product terms, and structured attributes. Combined with Query Type, that gives merchants a reasonably granular picture of where in a conversational journey their products appear — useful context if you’ve read our breakdown of how AI Mode surfaces brands as an information agent. What none of the axes give you is what happened after the impression.

The dashboard ships with four filters: product category, time period, countries, and traffic. Each carries a constraint worth knowing before you brief a stakeholder. The category filter works on a single category at a time — Google’s documentation states there isn’t a report that encompasses all categories. The countries filter is limited to eligible countries, which today means the US. And the traffic filter is the big one: current data is strictly limited to organic AI traffic, such as free listings. Paid Ads traffic isn’t included in this report.

The organic-only boundary deserves emphasis because Google is simultaneously building the paid side of the same surfaces — we covered AI Mode’s ad placements and user growth earlier this year. A merchant running Shopping ads into AI experiences will see none of that activity here. That’s not a flaw so much as a design decision, but it means AI performance insights can never be your single source of truth for AI-surface spend decisions — it measures the free half of a two-sided system.

No single piece of coverage lays out the full in-scope versus out-of-scope picture in one place — Search Engine Roundtable documents the metric list, Semrush frames the data views, Search Engine Land covers the May announcement. The table below is our synthesis of all three plus Google’s own help documentation: every signal the pilot report confirms, everything it explicitly excludes, and the gaps nothing documents yet.

Read as a whole, the scope map tells you what Google thinks merchants need first: reassurance that AI surfaces aren’t a black box. Every confirmed signal answers “am I visible?” — none answers “is it working?” That’s a rational sequencing for a pilot; impression data is the foundation everything else gets built on. But it also mirrors a pattern we’ve seen across 2026: AI shopping surfaces scale in quarters, while the measurement tooling that would let merchants manage them arrives in fragments, months later, region by region.

Our projection: the report’s obvious maturation path is clicks first, then some form of conversion linkage, then paid-organic unification — the same arc Search Console and the ads platforms walked years ago. If Google follows it, today’s pilot metrics become the baseline against which those later metrics get read. Which is precisely why getting access early, and archiving your Share of Voice trend from week one, may matter more than the numbers themselves.

The contrast between the surface and its measurement tool is the most telling number pair in this story. AI Mode reached general availability in the US in May 2025 and expanded to 180 countries by August 2025. By Google’s own I/O 2026 recap, it passed one billion monthly users within a year, and Google cites AI-feature reach of “nearly 200 countries and territories across 98 languages.” AI Overviews’ audience is even larger — widely cited at roughly 2.5 billion monthly users in secondary coverage, though Google’s own published I/O text doesn’t isolate that figure, so treat it as reported rather than Google-stated.

That is the trend worth interpreting: the gap between where AI shopping happens and where it can be measured is enormous, and it closes slowly. Merchants outside the US are selling into AI surfaces today with zero first-party visibility data and no committed date for getting any. The nearest proxy remains the Search Console side of the house — our guide to tracking AI Overview traffic in Search Console covers what’s observable there — but Search Console reports site-level search performance, not product-level shopping visibility. For catalog-level questions, this pilot is currently the only first-party game in town, and only in one town.

There’s no documented way to request pilot access, which makes the practical question: what should a merchant do so the report is useful on the day it appears? Four moves, in priority order — all of them valuable independent of the pilot.

The feed work is the highest-leverage move because it’s the only one that changes the number the report will eventually show. Our guide to preparing product feeds for AI shopping surfaces walks through the attribute-level detail. If you’d rather have a team audit your feed, competitor set, and measurement stack together, that’s the core of our ecommerce services — and for the reporting layer that ties AI visibility signals into the dashboards your leadership already reads, our analytics practice builds exactly that.