AI Performance Insights in Merchant Center: A Detailed Walkthrough

At Google Marketing Live on May 20, 2026, Google announced a new AI performance insights tool inside Merchant Center. The feature gives brands a view into how their products perform in shopping journeys originating from generative AI features in Search and Gemini, with share of voice as the primary metric.

The tool hasn't launched yet. Google says it's rolling out in Australia, Canada, India, New Zealand, and the US in the coming months. No exact dates, no European rollout confirmed yet. But the announcement itself is worth paying attention to, because Google released a detailed screencast showing what the reporting interface will look like, and the level of detail is more impressive than most marketeers were expecting.

Here's a walkthrough of what's coming, section by section, based on Google's own screencast.

The screencast Google shared

Google included a short screencast video in their announcement. It's worth watching in full before going into the individual sections:

For the full announcement, you can read Google's post here: How we're helping retailers thrive with new Universal Commerce Protocol features and AI tools on Google.

Let's go through what the screencast shows, section by section.

Section 1: your share of voice vs your competitors

The headline view of the AI performance insights tool shows your share of voice on Google's AI surfaces, compared to the average share of your competitors. The screencast example shows a brand at 14.5% share of voice (up 1.1%) against a competitor average of 11.2% (down 1.7%).

This is genuinely exciting, and worth understanding why. Merchant Center reporting covers performance data from across Google's full shelf space, which means this share of voice metric is coming straight from the source that actually serves the AI shopping responses. If you're familiar with how GMC reporting works, you'll know how valuable that level of data ownership is. For more on that, see Google Merchant Center clicks vs Search Console clicks.

This kind of data isn't available from any other AI platform. OpenAI doesn't tell you how your products perform in ChatGPT shopping responses. Claude doesn't expose this either. Google does, because Google owns both the surfaces (AI Mode, Gemini) and the merchant data (Merchant Center) on both sides of the equation.

The interface also lets you filter by product category, location, traffic type (organic in the example), and time period. The default in the screencast is the last 28 days.

Section 2: funnel performance breakdown

The next section breaks share of voice down by funnel stage. Google has defined three stages: Discovery, Evaluation, and Purchase. The screencast shows the Discovery section at 18.1% share of voice, with the Evaluation section visible underneath.

Each funnel stage maps to different query types:

• Discovery covers broad product searches by category, searches by product features, and searches by style.
• Evaluation covers queries that signal a shopper is comparing options.
• Purchase covers the queries closest to conversion.

This funnel concept maps closely to what marketeers usually call TOFO, MOFO, and BOFO (top, middle, and bottom of funnel). We wrote about why bottom of funnel visibility matters in detail here: why bottom of funnel visibility matters more than ever.

The funnel breakdown is potentially the most strategically useful part of the tool. A brand might have decent overall share of voice but lose almost all of it at the purchase stage, where intent is highest. Or the opposite: strong purchase-stage share of voice but weak presence during discovery, meaning shoppers only find you once they already know your brand. Seeing that split could change how brands prioritize their work.

Section 3: popular product terms

Below the funnel breakdown, the screencast shows a section called "Popular product terms." This lists the terms your brand is showing up for, each with its own popularity rating (shown as filled-in dots), the number of your products that apply, and a share of voice percentage.

The example in the screencast (for what looks like a shoe retailer) shows terms like "maximum cushioning," "plantar fasciitis," "arch support," "wide toe box," and "slip resistant." Each one shows how popular the term is, how many products in your catalog apply, and your current share of voice for that term.

The strategic value here is clear: spot terms with high popularity but low share of voice, and you've found an opportunity. In the example, "maximum cushioning" has 41.2% share of voice across 84 products. "Plantar fasciitis" has 35.4% across 18 products. Those are reasonable numbers. But "slip resistant" sits at just 4.7% share of voice with only 2 products applying. If slip resistance is genuinely relevant to more of the catalog, that's a clear cue to update product descriptions and reflect that attribute.

The call to action Google shows in this section is "Update product descriptions." That's Google explicitly telling brands: if you want to rank for these terms on AI surfaces, change your feed.

This is where supplemental feeds become genuinely useful. Rather than rebuilding your primary product feed every time you spot an opportunity, you can layer description optimizations on top using a supplemental feed. Productrise has a feature specifically for this. We cover how it works in detail here: how to use supplemental feeds to optimize organic shopping without touching paid campaigns.

Section 4: popular product attributes

The last section shown in the screencast is "Popular product attributes." This is similar to the product terms section, but focused on structured product attributes rather than free-text terms. The example shows attributes like color, style, material, fit, and size, each with a popularity rating and an attribute completeness percentage.

The completeness percentage is the part to focus on. In the example, color is at 82% completeness (most products have it filled in), but style is only at 44% and material at 42%. If those attributes are popular among shoppers asking AI assistants about products, missing them on more than half the catalog is leaving real visibility on the table.

This is another case where supplemental feeds shine. Rather than going back to your primary feed source and adding attributes one by one, you can use a supplemental feed to fill in the gaps. Productrise's feed audit checks for missing attributes today, and supplemental feed generation lets you add them at scale.

What's still unclear

A few things Google hasn't shared yet:

• Whether the data will be exposed through the Merchant API (likely yes, but not confirmed)
• How often the share of voice data refreshes
• Exactly how "competitors" gets defined behind the scenes
• Whether and when the tool will expand to European markets

The screencast gives a good preview of what to expect, but the final shipped product might look different. We'll update this post once the tool actually launches.

The new AI features rest on the same foundation as organic shopping

Worth emphasizing: AI performance insights is a new measurement tool, but the signals that determine whether your products surface in AI Mode and Gemini are the same signals that drive organic product visibility today. Clean product feed data, accurate structured data on product pages, relevant titles and descriptions, product reviews, and external mentions all matter.

Google didn't announce a separate ranking system for AI surfaces. It announced a way to measure performance on those surfaces. Every call to action in the screencast (update product descriptions, update product attributes) points back to the same fundamentals that already drive organic shopping rankings. If your feed is full of issues or your descriptions are thin, AI surfaces are going to reflect that.

For a deeper look at the basics, see how to audit your product feed for organic visibility and organic product SEO and bottom-funnel visibility.

How Productrise fits in

We connect to Merchant Center through the Merchant API. Once Google ships AI performance insights and exposes the data through the API, we'll be looking at how to bring it into Productrise in a useful way.

What Productrise already does today maps closely to what AI performance insights is showing:

• Tracking organic product visibility across queries (our carousel tracking)
• Measuring share of voice across categories (our query groups)
• Auditing product feeds for missing attributes and weak descriptions (our feed audit)
• Generating supplemental feeds to fill gaps in descriptions and attributes (our supplemental feed feature)

The CTAs Google is putting next to its AI performance insights (update product descriptions, update product attributes) are the exact use cases Productrise already supports through supplemental feeds. As soon as Google opens up the AI performance data through the API, we'll be working to bring it into the same place where marketeers already manage their feed optimizations.

Closing thoughts

AI performance insights is one of the more substantial Merchant Center announcements of recent months. It signals that Google considers AI Mode and Gemini real shopping surfaces worth measuring properly, and it gives brands the kind of data they can't get from any other AI platform.

The tool isn't live yet. But the screencast gives a clear preview of what's coming, and the optimization work it points to (richer descriptions, more complete attributes, better-mapped product terms) is work brands can start on today, regardless of whether their market is in the first rollout wave.

Want to start optimizing your descriptions and attributes now, so you're ready when AI performance insights launches? You can get started with Productrise for free. No credit card required.