Google Merchant Center AI Performance Insights: An India Audit Guide

What changed for Indian merchants in September 2026?

Google first previewed the report for India earlier in 2026. Its 16 September commerce update changed the status from a future rollout to current availability across India and four other markets.

The release sits beside product-feed and agentic-commerce updates, but not every feature has the same geography. AI Performance Insights is available in India; the Business Agent for YouTube ads beta is US-only; and new Universal Commerce Protocol checkout capabilities are initially rolling out in other markets. Record the availability of each feature separately rather than treating the announcement as one global launch.

Google's India Marketing Live summary also positions AI Performance Insights as the visibility layer for AI Mode, AI Overviews and Gemini-era shopping. The current Help Center page, however, defines the report's live scope as AI Mode and AI Overviews. Use the product documentation visible in the account as the operational source of truth.

How should AI share of voice be interpreted?

Google calculates your share using your AI impressions divided by the combined AI impressions for your brand and the Merchant Center competitor set on related queries.

That makes it a relative visibility metric. It can move because:

• your products appeared more or less often;
• a competitor's visibility changed;
• Google changed the available competitor set;
• query demand shifted within the selected category;
• product eligibility, stock or data quality changed; or
• the time period includes reporting lag.

Google says merchants cannot choose the comparison brands. It also notes that your share and the competitor average can rise together when the competitor data changes. A 100% share can mean insufficient competitor data, not category dominance.

Report share of voice with at least three companion measures:

| Metric | What it tells you | What it does not prove | | --- | --- | --- | | AI share of voice | Relative organic visibility for the selected scope | Clicks, sales or incrementality | | Frequency | How common a term, intent or attribute is in the report | Demand available specifically to your brand | | Products showing | Breadth of eligible products appearing | Product quality or conversion readiness | | Organic sessions | Visits measured by your analytics implementation | All AI impressions or assisted influence | | Net orders and contribution | Commercial outcomes under your finance rules | Outcomes caused by AI visibility alone |

The AEO glossary and HML's AI-search measurement guide explain why visibility, citations, referrals and conversions require separate measurement layers.

Which report views should a D2C team audit first?

Which category and country are being measured?

Google requires a single product-category view; there is no all-category AI Performance report. Select India, one category and a stable time range. Save the exact filters with every export or screenshot so the next review compares like with like.

Where is visibility weak across the shopping journey?

Compare discovery, evaluation and ready-to-buy share. A brand may appear in broad inspiration queries but disappear when shoppers ask about specifications, alternatives, compatibility or reviews. That gap suggests a product-information problem, not necessarily a media problem.

Which high-frequency terms have low coverage?

Prioritise terms where frequency is high, share is weak and few relevant products show. Then verify that the term describes a real product fact or use case. Google's own guidance recommends incorporating relevant terms into titles and descriptions, but relevance must come before coverage.

Which popular attributes are missing?

Attributes such as material, size, compatibility, ingredients or technical specifications help systems answer comparison questions. Google's product detail documentation supports clean section, attribute and value pairs for product-specific information. It explicitly advises against listing keywords in place of genuine specifications.

How should product data be improved without keyword stuffing?

Use the AI report as a diagnostic queue, not an instruction to copy every phrase into every title.

1. Fix eligibility first. Resolve disapprovals, policy issues and missing required attributes.
2. Stabilise identity. Check brand, GTIN or MPN, item group, variant relationships and canonical product URLs.
3. Reconcile offer facts. Match price, currency, availability, condition, delivery and returns across the feed and landing page.
4. Add discriminating attributes. Submit factual material, size, colour, pattern, capacity, compatibility, ingredient or specification values that help a shopper compare.
5. Improve descriptions. Explain the product's use, salient features and limitations in natural language without promotional filler.
6. Validate structured data. Align Product and Offer markup with the same values provided to Merchant Center.

Google's product data specification warns that missing identifiers, variant attributes, images or conflicting website data can restrict eligibility. Its structured-data mapping guide recommends matching schema.org properties with Merchant Center values, especially price, currency, availability and condition.

HML's Shopify agentic-storefront guide covers the wider catalogue architecture. Use the AEO Readiness Checker to find answer and entity gaps on the landing page after feed defects are addressed.

What does a practical AI visibility audit look like?

Consider a fictional Indian footwear brand reviewing one category for two mature 28-day windows. The values are synthetic and demonstrate the reporting logic, not a market benchmark.

| Measure | Baseline | After remediation | Interpretation | | --- | ---: | ---: | --- | | AI share of voice | 8% | 12% | Relative visibility improved within Google's comparison set | | High-frequency attributes complete | 54% | 86% | More relevant product facts were supplied | | Products showing for top terms | 14 | 29 | Eligible catalogue coverage widened | | Tagged organic AI sessions | 180 | 245 | Observed visits increased, subject to attribution limits | | Net orders from tagged sessions | 9 | 13 | Commercial outcome improved, but sample is small | | Incremental orders | Not tested | Not tested | Causality remains unknown |

Method: keep the country, product category and reporting window length constant; archive the selected competitor and query views; log every feed and landing-page change; exclude out-of-stock products from the remediation denominator; preserve transaction IDs through analytics and finance; and wait for reporting lag before comparing mature windows.

The correct conclusion is that visibility, data completeness and observed visits improved together. It is not defensible to say that the four additional net orders were caused by the feed changes without a credible counterfactual.

How should AI visibility connect to revenue reporting?

Create four reporting rows:

1. Merchant Center visibility: share of voice, frequency, products showing and shopping stage.
2. Site behaviour: tagged sessions, product views, engaged visits, add-to-cart and checkout starts.
3. Finance outcomes: net orders, new customers, contribution and returns under a disclosed rule.
4. Causal evidence: controlled lift when the scale and decision justify a test.

If AI visibility improves but sessions do not, inspect whether users can click through from the relevant experience and whether analytics preserves the source. If sessions rise but conversion does not, inspect price, availability, delivery, trust and landing-page continuity. If platform and finance totals disagree, resolve transaction IDs and order-status rules before declaring growth.

The attribution glossary is useful when teams are tempted to assign every assisted outcome to a single surface. HML's Shopify analytics case study provides implementation context for making product and transaction data auditable; its results are not evidence for the synthetic example above.

What should the Merchant Center owner do this week?

Build a one-page AI visibility register for each priority category:

• selected country, language, category and time period;
• current share of voice and competitor average;
• high-frequency terms with low share;
• popular attributes missing from relevant products;
• products showing versus eligible catalogue size;
• feed, landing-page and structured-data owner;
• remediation date and validation status; and
• analytics and finance outcome rows.

Rank fixes by frequency × commercial relevance × affected products, then subtract implementation risk. A missing material value across 200 best-selling products usually deserves attention before a low-frequency phrase affecting two products.

If the audit reveals conflicts across Merchant Center, Shopify, structured data and analytics, request a product-data and AI-visibility audit or review HML's website development service. The intended deliverable is a field-level remediation backlog tied to measurable shopping stages, not a generic promise to “optimise for AI.”

Sources

1. Boost your holiday sales with these agentic commerce updates — Google Shopping Blog; published 16 September 2026; accessed 19 September 2026.
2. About AI performance insights — Google Merchant Center Help; accessed 19 September 2026.
3. Product data specification — Google Merchant Center Help; accessed 19 September 2026.
4. [Product detail [product_detail]](https://support.google.com/merchants/answer/9218260) — Google Merchant Center Help; accessed 19 September 2026.
5. Supported structured data attributes and values — Google Merchant Center Help; accessed 19 September 2026.
6. Google Marketing Live 2026: Delivering the Gemini advantage for Indian businesses — Google India Blog; accessed 19 September 2026.