Google AI Performance Insights: What 0% and 100% Share of Voice Really Mean

A 0% AI share of voice is not a diagnosis of a broken catalog. A 100% share is not proof that your brand owns the category.

Google's AI performance insights can now help ecommerce teams see how their products appear in conversational shopping searches. The useful decision is where to investigate and improve product data, not whether to celebrate or rewrite the catalog based on one percentage.

Start with the number beside share of voice: Products showing. Then check the report's scope and competitor data. Those checks separate a visibility problem worth investigating from a reporting condition that more product copy will not fix.

What changed in September

Google announced general availability on September 16, 2026 for businesses in Australia, Canada, India, New Zealand and the United States. This is the rollout of the tool introduced at Google Marketing Live in May, not a second new measurement product.

The current Merchant Center documentation defines its scope as English-language conversational queries with shopping intent on AI Mode and AI Overviews. The report covers organic AI traffic, including free listings, rather than paid ads. It is not a readout of the Gemini app or other AI shopping platforms.

To find it, open Merchant Center and go to Analytics, then Products, then AI performance. Select a product category, country and time period. There is no all-category report, so label each view before sharing it with your team.

Read these four results differently

Google's troubleshooting guidance makes distinctions that are easy to lose in a dashboard screenshot.

0% share of voice

Google says insufficient impressions can display as zero.

Read Products showing and check the selected scope before treating this as evidence that none of your products appear.

A dash for share of voice

Google says a dash means no impressions data.

Treat the view as missing evidence, not a measured competitive loss. Check another relevant period or scope.

0 products showing

No products showing in that view. Google says this is an exact count.

Investigate the relevant products and submitted data for that scope. The count establishes absence, not its cause.

100% share of voice

An account without defined Merchant Center competitors displays 100%.

Check whether competitor data is available before reporting category leadership.

The distinction matters when approving work. A zero share-of-voice value with products showing is a different starting point from zero products showing. Neither justifies a catalog-wide enrichment project on its own.

The denominator is a competitor set, not the whole market

Google calculates share of voice from your AI impressions divided by the combined impressions for your brand and its defined Merchant Center competitors on related queries.

That makes it a relative visibility measure within a defined comparison, not your percentage of all AI shopping, all category demand or ecommerce revenue.

You cannot change the competitors in this report. Google also says your own share-of-voice average and your competitors' average can rise at the same time when competitor data changes, including changes to the set. A rising percentage therefore deserves a scope check before it becomes a performance claim.

For a useful trend, keep the category, country, time period length and shopping stage consistent. Note any visible changes in competitor data. The report updates daily with a few days of lag, so do not use the current reading to grade a feed change made the same morning.

Use the report to choose a specific product-data fix

The most actionable view combines shopper demand, your visibility and the products involved.

Google separates discovery, evaluation and ready-to-buy queries. It also provides top terms, popular attributes and top search intents. Frequency indicates how popular those terms, attributes or intents are. Products showing counts your products appearing for them. Share of voice is not available in the popular attributes scorecard.

Start with a high-frequency need that genuinely matches products you sell. Then inspect the relevant submitted product data.

For example, suppose a running-shoe team sees “arch support” among its top terms and weak visibility for feature-led discovery. That is a reason to check whether suitable shoes have accurate support information in their product data. It is not a reason to add “arch support” to every shoe.

A useful first pass looks like this:

1. Choose one category and one shopper need. Save the report's scope and the current readings.
2. Identify matching products. Use their specifications, not the popularity of the term, to decide which products belong.
3. Check what Google receives. Compare the submitted titles, descriptions and relevant attributes with the actual product facts. Fix missing or inaccurate information rather than adding synonyms indiscriminately.
4. Record the exact change. Keep the affected product list and submission date so the next review has something concrete to compare.
5. Review after the reporting lag. Check products showing and the applicable visibility metrics in the same scope. Track traffic and commercial outcomes separately.

Google recommends starting with popular relevant terms and missing attributes. The discipline is relevance: a demand signal tells you what shoppers want, not what your product can truthfully claim.

Conversational attributes are useful when they add information

The rollout is not a mandate to rebuild every feed.

Google's conversational attributes guidance says these fields are optional and complement existing product data. Adding them does not change existing product approval status. If you already submit a detail in the description, product highlight or product detail fields, Google says you do not need to duplicate it in conversational attributes.

Use the fields for information that helps a shopper make a decision: a product-specific question and answer, a manual, a related accessory or clearer variant information. The task is to make useful facts available, not to fill a new column for its own sake.

Google's September announcement gives a bounded example. During testing with lululemon, submitted conversational attributes were incorporated 50% of the time in relevant product recommendations in AI Mode. That is evidence of Google using the information. It is not a 50% increase in recommendations, traffic or sales.

Keep visibility and business results on separate lines

For an ecommerce review, report three things separately:

• Visibility: What appeared in Google's measured scope, and how did share of voice compare with the defined competitor set?
• Product-data work: Which factual gaps did the team confirm and correct?
• Commercial results: What happened to measured visits, orders and revenue in your analytics and commerce reporting?

The “ready to buy” stage describes query intent. It is not a count of purchases. Likewise, a higher share of voice does not establish that your data edit caused incremental sales.

If another AI platform matters to your business, measure that platform separately. Do not combine Google's impression-based share with a manual prompt sample and label the result one universal AI visibility score.

Before approving the next enrichment project, ask for one report view, one verified product-data gap and one defined change. Google's report can help you choose that work. The percentage alone cannot choose it for you.

Google's report tells you where to look. The work that follows is checking the submitted titles, descriptions and attributes against the product facts and fixing what is wrong. That is how Paz works it for lean ecommerce teams: a Paz expert finds the gaps in your Merchant Center product data, you approve the fixes, and the platform publishes them and tracks share of voice and products showing in the same scope. Talk to us.