Google AI Overviews and Product Discovery: What Ecommerce Managers Need to Know
Google’s AI Overviews have arrived in product search, and they are fundamentally different from organic search results and Shopping ads. They are not ranked links. They are generated answers, with embedded product recommendations, that appear before any other search result. For ecommerce managers, the question is not whether AI Overviews will affect your traffic. They already do. The question is what determines whether your products appear in them, and what you need to change to be consistently included.
How AI Overviews Generate Product Recommendations
AI Overviews are not a new ad format or a new search result type. They are a generative AI response layer built on top of Google’s existing knowledge infrastructure, primarily the Shopping Graph and the Knowledge Graph. When a shopper asks a shopping-intent question (“what’s a good sustainable hiking jacket for UK weather under £150?”), Google’s AI system does not retrieve a list of keyword-matched web pages. It generates a structured response that synthesizes information from multiple sources and embeds specific product recommendations pulled from the Shopping Graph.
The mechanism for those product recommendations is not keyword matching. It is structured attribute filtering against the Shopping Graph, the same mechanism an AI agent uses. Google’s AI system translates the query into evaluable criteria:
Products in the Shopping Graph that have these attributes populated as structured fields and that meet the criteria appear in the AI Overview. Products that have “sustainable” mentioned in a description paragraph but not in a structured attribute field may not.
AI Overviews Are a Structured Query Engine, Not a Search Engine
This is the most important reframe for understanding AI Overview inclusion. Traditional SEO optimizes for keyword matching in a probabilistic ranking system. AI Overview optimization requires structured attribute completeness in a deterministic filter system. The SEO playbook, keyword density, title tags, meta descriptions, and link building, has minimal direct effect on AI Overview product inclusion. Attribute completeness, GTIN entity matching, and precise product_details pairs are the variables that matter.
The Shopping Graph: The Database Behind AI Overviews
Every product recommendation in a Google AI Overview is served from the Shopping Graph, a product knowledge base containing over 35 billion product listings that Google maintains, enriches, and continuously updates. Your product’s representation in the Shopping Graph is built from three sources:
- Your Merchant Center feed, the data you explicitly submit, the primary data input.
- Google’s crawl of your product pages, the data Google’s bot discovers from your PDP HTML and schema.org markup.
- The Knowledge Graph, Google’s pre-existing entity database, activated by GTIN entity matching.
The quality and completeness of your Shopping Graph entity, the composite representation Google maintains for your product, determines your eligibility and ranking in AI Overview recommendations. A product with a rich, complete Shopping Graph entity, strong GTIN entity match, complete product_details pairs, high-quality imagery, accurate pricing, aggregated reviews, is both more likely to be retrieved and more likely to be recommended for relevant queries.
How a Shopping Graph entity is built
Merchant Center feed
Your submitted data, especially product_details, forms the primary input.
Page crawl + schema
Google independently reads your PDP HTML and schema.org markup.
Knowledge Graph entity
GTIN matching strengthens identity, confidence, and inclusion probability.
The 4 Data Factors That Most Influence AI Overview Inclusion
product_details — The Most Critical Feed Field for AI Overviews
product_details is the feed attribute that accepts structured specification pairs, attribute name / attribute value, for any technical or descriptive product feature. It is the field that most directly feeds AI Overview structured attribute matching. When Google’s AI system evaluates whether your product matches “sustainable materials” or “waterproof for UK weather,” it reads product_details pairs, not your description text.
Examples of product_details pairs for a hiking jacket:
Waterproof Rating / 20,000mm HH | Weight / 490g | Material / 100% Recycled Polyester | Packable / Yes | Sustainability / Bluesign Certified, OEKO-TEX Standard 100 | Hood Type / Adjustable, Removable
Each of these pairs is a structured matching target for AI Overview queries that include the corresponding criterion. Products without these pairs cannot match those queries, even if the information exists in the product description.
GTIN Entity Matching Unlocks Elevated Inclusion
Products with valid GTINs that match a Knowledge Graph entity receive elevated treatment in the Shopping Graph. Google’s AI system has higher confidence in entity-matched products. It has independent verification of the product’s category, brand, and basic specifications from the Knowledge Graph, supplementing what you submit in the feed. Entity-matched products are also eligible for cross-merchant comparison surfaces that are embedded in AI Overviews, the price comparison and “available from multiple sellers” formats that appear within AI-generated product recommendations. These surfaces are inaccessible to non-entity-matched products, representing a qualitatively different level of visibility.
Schema.org and Feed Agreement — The Trust Signal
Google reads your product page’s schema.org markup as an independent data source separate from your Merchant Center feed. When schema markup and feed data agree on price, availability, product specifications, and brand, Google’s confidence in both signals increases, which positively affects your Shopping Graph entity quality and, by extension, your AI Overview inclusion probability. When they disagree, schema says £89.99, feed says £79.99, Google detects a conflict. The conflict reduces trust in both data sources. In an AI Overview context, lower trust means lower inclusion probability, because Google’s AI system is making a recommendation to a consumer and cannot recommend a product whose data signals are contradictory.
Review Data via Schema — The Quality Signal
AI Overviews include social proof signals alongside product recommendations, typically star rating and review count. This data is pulled from aggregateRating schema on your product pages and from the Shopping Graph’s cross-merchant review aggregation for entity-matched products. Products without aggregateRating schema implemented, even if they have strong reviews on their website, may not have review data available for the AI Overview recommendation format. Implementing review schema is a one-time technical task that makes your review data visible to AI systems that read schema, not just the human shoppers who see it on your page.
The Gemini Shopping Angle: Conversational Product Discovery
Google Gemini’s shopping integration represents the most conversational and criteria-rich form of AI product discovery currently available. A Gemini query like “I need a jacket that can handle Scottish Highland weather in October, weighs under 600g, and costs less than £200, what do you recommend?” is parsed into a multi-criteria filter against the Shopping Graph that simultaneously evaluates weatherproofing, weight, price, and contextual suitability.
The products that appear in Gemini shopping recommendations are those with the most complete and most precisely specified attribute data in the Shopping Graph for the query’s evaluated criteria. This is not speculative. You can test it directly. Search for your own product category in Google with a multi-criteria specification query. The products that appear in any AI Overview or Gemini recommendation are the benchmark for what complete attribute data looks like in your category.
The Competitive Intelligence Technique
Search for your top 5 product categories using conversational, multi-criteria queries in Google. Observe which products appear in AI Overviews and Gemini recommendations. Those products have the attribute completeness that is currently sufficient for AI Overview inclusion. Pull their product_details feed data, visible in Google’s Shopping product panels, and use it as your enrichment benchmark. Your target is to match or exceed their attribute coverage for every criterion that appears in natural-language queries for your category.
Google AI Overviews Enrichment Checklist
Inclusion driver
Structured product_details pairs are the main route into matching, not better keyword copy.
Confidence driver
GTIN matching and schema-feed agreement increase trust in the entity Google builds for your product.
Presentation driver
Review schema, highlights, and precise attributes help the AI Overview recommendation look stronger once included.
Velou on the AI Overviews Opportunity
AI Overviews represent the most significant product discovery surface change in Google Search since Shopping ads launched. The competition for inclusion is still forming. The retailers investing in product_details completeness, GTIN entity matching, and additionalProperty schema now are building inclusion probability advantages that will be difficult for later movers to close.
Commerce-1’s Google enrichment mode generates product_details pairs, product_highlights bullets, and schema additionalProperty fields simultaneously, the full data package that AI Overview inclusion requires.
Build the data foundation for Google AI Overview inclusion
Commerce-1 generates product_details, highlights, and schema simultaneously, the full AI Overview data package.
Request a demo
.png)
.png)