4 ways online stores should use AI performance insights in Merchant Center

AI performance insights is a new report in Google Merchant Center that shows how your products are discovered in AI Mode, AI Overviews and the Gemini app. Online stores should use it in four ways: to benchmark share of voice against competitors, to map visibility across the shopping journey, to rewrite product content around the terms shoppers use, and to close gaps in product attributes.

The feature has been announced but is not yet live in most accounts. That gap is the opportunity. Every step in this post can start today with the data you already have, so your product feeds are ready when the AI performance tab appears.

What are AI performance insights in Merchant Center?

Google is introducing AI performance insights as an analytics report for organic visibility on its conversational shopping surfaces. It sits in Merchant Center under Analytics, then Products, then a new AI performance tab at the top of the page.

The report is built to focus on conversational search queries. These are the longer, question-style prompts people type into AI Mode, such as "a waterproof hiking jacket that packs small for under 200 dollars." Classic Shopping reports were built around short keywords. This one is built around intent and context.

In Google's words, the report helps you understand how your brand shows up in conversational queries, evaluate your visibility across shopping stages and view trends to help optimize your product info.

Google's help documentation describes four main metrics:

• Share of voice: the share of voice captured by your brand or products compared with your competitors, based on AI impressions.
• Competitors' average share: the share of voice captured by the competitor set defined in your Merchant Center account.
• Frequency: how popular a search type, search term, search intent or attribute has been.
• Products showing: how many of your products appear for top terms, popular attributes and search intents.

On top of those metrics, the report breaks visibility into shopping stages and surfaces the product terms and product attributes shoppers ask about most. Together, these views are meant to help brands see where they stand in AI shopping, not only whether they appear.

‍

When will AI performance insights be available?

Google first presented the feature at Google Marketing Live in May 2026. In its shopping updates announcement, Google said the tool "gives you a clear view into how your brand is performing on AI surfaces by comparing your share against similar brands."

Google says it will roll out to the United States, Canada, Australia, India and New Zealand in the coming months. The help page limits the report to English-language queries for Merchant Center accounts in those markets.

In July 2026, Search Engine Roundtable reported that the report was in a limited pilot with select Merchant Center accounts in the US. Hana Kobzova of PPC News Feed first spotted the new help documentation. So far, only a small number of Merchant Center accounts can see the tab. European accounts are not part of the announced rollout at all.

How AI shopping changes search visibility

Search visibility used to mean ranking for a keyword. A shopper typed "running shoes," saw a grid of Shopping ads and free listings, and compared them on the results page.

Google AI changes that pattern. In AI Mode, a shopper can describe a need in full sentences and refine it through a conversation. The answer blends product listings, reviews and explanations into a single response. Fewer products make it into that response, and the ones that do are picked because their information matches the details in the prompt.

This has three practical effects. First, consumer demand shows up as intent, not keywords. You can identify the nature of that demand in the AI performance insights report, sorted by shopping stage and product category. People still search by category, but they add context such as budget, use case or a problem to solve. Second, the product data you send decides whether your products surface at all. Third, the metrics you used to track Shopping performance do not tell you how you appear in conversational results.

That last point is the gap AI performance insights is designed to fill. Until now, merchants had no general view of AI visibility inside Google's own tools. The new AI performance insights give a first, official measure of how AI experiences present your brand across these new shopping experiences.

Why preparing early gives you an advantage

When the report lands, it will show a baseline. If your product data is thin on that day, your first share of voice number will be low, and you will spend weeks catching up while competitors who prepared already hold the position.

AI surfaces also reward completeness in a way keyword bidding does not. The general rule is straightforward: the more structured and accurate your product data, the more likely your products are to surface in a relevant conversational answer. A conversational answer pulls the products whose data matches the details in the prompt. A feed that only passes Merchant Center validation is not the same as a feed that answers questions about fit, material, compatibility or use case.

So the first-mover advantage here is not about being first to open a dashboard. It is about having a feed that already performs well on the metrics the dashboard will measure. Whatever happens with the rollout date, that work improves your free listings and Shopping ads too.

1. Benchmark your share of voice against competitors

Share of voice is the headline metric. Google calculates it as your AI impressions divided by the total impressions for you and your competitors on related queries.

Check your competitor set now

The report uses the competitor set defined in your Merchant Center settings. According to Google's documentation, you cannot change these competitors manually inside the report. That makes it worth reviewing now which brands Merchant Center already treats as your competitors in existing reports such as price competitiveness and best sellers.

If the set looks wrong, the cause is usually your product category data. Check that your google_product_category and product_type values describe what you sell as precisely as possible. A vague product category places you against the wrong brands.

Build a manual baseline today

You do not need the report to get a rough picture. Pick 20 to 30 conversational prompts that match your best-selling categories. Run them in AI Mode from a supported country, or ask a colleague there to do it. Note which brands appear, which products are shown and what details the answer highlights.

Repeat the exercise each month. When the official share of voice data arrives, you will be able to compare it with your own record and see whether the changes you made moved the numbers.

Track changes and spot patterns

Once the report is live, track your share of voice captured by week rather than by day. Look for the categories where your number moves against the competitor average. A sudden drop often points to a feed issue, such as disapproved products or a price change. A slow decline usually means competitors have improved their product data faster than you.

Use the frequency metric alongside share of voice. A low share on a rare query matters less than a low share on a popular one. This helps you identify where effort will pay off first.

Read the numbers carefully when they arrive

Google notes two edge cases. A share of voice of "0" means there were not enough impressions to report a value. A dash means there is no impression data. Neither means your products are invisible, so check the frequency metric before you draw conclusions.

2. Map visibility across discovery, evaluation and ready to buy

The report shows visibility across different phases of the shopping journey. It splits conversational queries into three stages. Discovery covers exploration, when someone is still working out what they need. Evaluation covers comparison and specification questions. Ready to buy covers queries with clear intent to purchase.

Most stores are strong at one stage and weak at the others. A brand with good prices may show up well for ready to buy queries but disappear during evaluation, when shoppers ask how products compare. A brand with rich editorial content may appear during discovery but lose the sale because its offers lack shipping or availability data.

Match content to each stage

For discovery, your product descriptions need to explain what the product is for and who it suits. Use-case language matters here. "A carry-on backpack for weekend city trips" gives an AI answer more to work with than a list of dimensions alone.

For evaluation, the details that drive comparisons need to be in the feed as structured data. Think size, weight, material, capacity, compatibility and certifications. If a shopper asks which option is lighter, the product with a weight value can be compared. The one without cannot.

For ready to buy, the essentials are price, availability, shipping and returns. Keep them accurate and consistent with your landing pages. Mismatches can trigger disapprovals and cost you visibility at the moment it matters most.

Plan your fixes by stage

Once the report is live, look for the stage where your share of voice drops furthest below the competitor average. That is where to focus first. Until then, audit your top 50 products against the three stages above and note which data is missing for each.

3. Use product term insights to rewrite titles and descriptions

The product term insights show which terms shoppers use in Search conversations, and your share of voice for each. Google's own examples are buyer-focused phrases such as "easy setup" or "long battery life."

These are not traditional keywords. They describe benefits and outcomes. A shopper in AI Mode is less likely to type a model number and more likely to describe a problem.

Find the terms before the report does

You can start collecting these terms now from sources you already own:

• Customer reviews on your site and marketplaces
• Questions sent to customer service
• Search terms reports in Google Ads, filtered for longer queries
• On-site search logs
• Return reasons, which often reveal expectations the product page did not set

Look for phrases that repeat. If many reviews mention that a blender is "quiet enough to use early in the morning," that is a product term worth reflecting in your description. Language driven by real user behaviour tends to match the way people phrase prompts in AI Mode.

Rewrite, do not stuff

Add these terms where they are true and useful. A title should still lead with brand, product type and key attributes. The description is the place for benefit language and use cases. When you write for conversational queries, write clearly for a person, not by repeating phrases for a machine.

When the report arrives, identify the product terms with high frequency and compare them with your list. Terms with high frequency and low share of voice are your fastest wins.

4. Close product attribute gaps with conversational attributes

The product attributes insights show the specifications shoppers search for, such as color, style and material. The report also flags popular attributes that may be missing from your product data.

This is where most feeds fall short. A product can be approved in Merchant Center with only the required fields filled. That is enough to run ads. It is often not enough to be picked for a detailed conversational answer.

Fill the recommended attributes first

Start with the optional attributes that already exist in Google's product data specification: color, size, material, pattern, age group, gender and product detail. For many categories these fields are left empty, even though they are exactly what AI answers compare.

Use product_detail and product_highlight for specifications that have no dedicated field. Keep values factual and consistent across variants.

Add the new conversational attributes

At the same event, Google introduced conversational attributes, which are available globally rather than in five countries. Search Engine Land covered the launch alongside AI performance insights. Google's help documentation lists six of them:

• question_and_answer: question and answer pairs about the product, such as whether it supports Bluetooth.
• document_link: links to related PDFs, such as manuals or assembly instructions.
• related_product: links to accessories, required parts and other related items, with the type of relationship stated.
• item_group_title: a single title for a product that comes in several variants.
• variant_option: the properties that distinguish variants, such as size and color.
• popularity_rank: a percentage that indicates how popular the product is.

The question_and_answer attribute is the most direct link to conversational search. Use the customer questions you collected in step three. Answer each one in plain, accurate language. Because these attributes are available worldwide, European stores can add them now, even though the AI performance tab is not planned for their accounts yet.

How to manage the report across several Merchant Center accounts

Many online stores run more than one Merchant Center account, often one per country or one per brand. At launch, only accounts in the five supported markets will see the report, and only for English-language queries.

If you manage a multi-country setup, map which of your accounts target those markets. A store based in France that sells into Canada or Australia may have product feeds that qualify, while its home-market feed does not. Check each account once the rollout reaches you, rather than assuming the report is missing everywhere.

It also helps to decide early who will access the data. The report lives in the Analytics section, so make sure the right user on your team has permission to view it. Agencies should confirm access for each client account in advance.

Finally, apply what you learn in one market to the others. Missing attributes and product terms found in an English-language account often point to the same gaps in your French, German or Spanish feeds. That is an important factor when deciding where to invest feed improvement work. Google may expand the report to more markets later, and those feeds will then be ready.

Limits to keep in mind

AI performance insights has clear boundaries, and knowing them prevents bad decisions.

The insights only focus on organic AI traffic. Search Engine Roundtable noted that traffic from paid ads is not included, so the report will not tell you how Performance Max or Shopping ads perform in AI Mode. On the other hand, that makes it a clean measure of how well your product data works without paid support. The report can expand your understanding of organic AI reach in a way no existing Shopping report does.

The data updates daily, but with a lag of a few days. It suits trend analysis more than day-to-day reactions.

Availability is limited to English-language queries in five countries. Your other storefronts will need other ways to monitor AI visibility for now.

The competitor set is fixed by Merchant Center. Your benchmark is only as good as your categorization.

Last, the report shows visibility, not revenue. It does not happen to include any conversion or revenue data. Share of voice is one factor in AI shopping performance. Keep measuring clicks, conversions and margin in your usual tools.

A 90-day preparation plan

In the first month, review your competitor set and product categories. Run your manual prompt baseline in AI Mode and save the results.

In the second month, audit your top products for missing attributes. Fill color, size, material and product detail where they are empty. Collect product terms from reviews, support tickets and search logs.

In the third month, rewrite descriptions for your priority categories around those terms. Add question_and_answer and related_product values to your best sellers. Run your prompt baseline again and compare it with month one.

By the time the AI performance tab appears in your account, you will have a cleaner feed, a baseline to compare against and a list of changes to measure.

Frequently asked questions

Is AI performance insights available in Europe?

No. Google has announced the rollout for the United States, Canada, Australia, India and New Zealand only, for English-language queries. There is no announced date for European markets.

Where will I find the report in Merchant Center?

Go to the Analytics tab, select Products and open the AI performance tab at the top of the page. If you do not see the tab, your account is not yet part of the rollout.

Does the report include Google Ads data?

No. It covers organic visibility on AI surfaces. Paid ads traffic is excluded.

Which AI surfaces does the report cover?

Google names AI Mode, AI Overviews in Search and the Gemini app as the surfaces the report is designed to cover.

Can I choose my own competitors?

Not inside the report. It uses the competitor set defined in your Merchant Center settings, which depends heavily on how your products are categorized.

How often is the data updated?

Daily, with a lag of a few days. Use weekly or monthly views for decisions.

What can I do before the report reaches my account?

Complete your product attributes, add conversational attributes such as question_and_answer, rewrite descriptions around real customer language and record a manual baseline of your visibility in AI Mode.

‍