If you sell on Google Shopping, Merchant Center now has a dashboard that answers a question you couldn’t measure before: how often do your products actually show up when someone shops through Google’s AI surfaces — AI Mode, AI Overviews, and the Gemini app?

What shipped

Google announced AI Performance Insights and a companion feature, Conversational Attributes, at Google Marketing Live on May 20, 2026. The pilot went live for a limited set of US Merchant Center accounts around July 13–14, and Google says broader rollout to Canada, Australia, India and New Zealand is coming “in the coming months” (Search Engine Land; Google Merchant Center Help).

Both features live inside the product-feed pipeline you already use — there’s no new file, endpoint or protocol to implement. It’s a data-completeness and measurement play, similar to how Search Console formalized the sitemap as a ranking input.

The four things the report shows

Found under Analytics → Products → AI performance in Merchant Center (Google Merchant Center Help; MarTech):

  • Share of voice — your AI-surface impressions divided by total impressions across you and a benchmark set of similar competitors, for the same queries.
  • Shopping funnel performance — discovery, evaluation and purchase stages, broken out specifically for AI-surface traffic rather than blended with Search.
  • Product term insights — the actual conversational queries shoppers used to find a listing (Google calls out query frequency and query type as sub-metrics).
  • Product attribute gaps — a completeness score flagging which structured fields (color, material, style, etc.) are missing per product, since Google’s AI systems need that structured data to match a query like “waterproof hiking boots under $150” to the right listing.

Conversational Attributes is the fix for that last item: merchants can now add free-form conversational attributes and richer descriptions directly in Merchant Center, which feed the same AI matching systems. It’s rolling out globally, ahead of the insights report itself.

What it doesn’t cover

This is a Google-only dashboard. It reports on AI Overviews, AI Mode and Gemini — not ChatGPT Shopping, Perplexity, Amazon Rufus, or any other assistant a shopper might use. A merchant who only watches this report gets a real signal, but a partial one; multi-surface visibility still means stitching this together with third-party AI-citation tracking elsewhere.

It also measures a different layer than what AgentReady checks. Share of voice and attribute completeness are discovery metrics — whether an AI surface shows your product. They say nothing about whether an agent that finds you can actually resolve price and stock and complete checkout, which is the harder, under-measured half of agent-readiness. A product with a perfect attribute-completeness score can still sit behind a cart an agent can’t reach.

What to do with it

The attribute-gap view is the actionable one: it’s a concrete, per-product checklist of missing structured fields, which is the same kind of gap a product-feed or schema audit should already be catching. If your account gets pilot access, treat a low attribute-completeness score as a to-do list, not a grade — and don’t mistake a strong share-of-voice number for evidence that a shopping agent can actually buy from you. Those are two different questions, and right now only one of them has a Google-native dashboard.

Sources

A dashboard that shows your product got cited is not the same as one that shows an agent could check out — AgentReady audits the second question.