The Shopping Graph is the data set Google uses to understand what is for sale online and in physical shops: which products exist, who sells them, at what price, whether they are in stock, and what reviewers and buyers say about them. Google assembles it with machine learning from merchant submissions and from material it finds across the web, then links those records so that one product sold by many retailers can be recognised as the same item. It exists because a search engine that answers shopping queries needs more than a list of web pages. It needs a structured, constantly refreshed map of products and how they relate.

How the graph is assembled

Google has never published a technical specification for the Shopping Graph. When Bill Ready, then Google's president of commerce and payments, introduced it on May 18, 2021, he described "a dynamic, AI-enhanced model that understands a constantly-changing set of products, sellers, brands, reviews, product information and inventory data", along with how those attributes relate to one another, according to Google's announcement.

Four broad inputs feed it. The first, and the most structured, is the merchant product feed: the file or API stream a retailer submits through Google Merchant Center, carrying identifiers, titles, prices, images, availability and barcodes. Google's help page on sources of shopping information names Merchant Center and Manufacturer Center as the tools through which brands and retailers "send their product info directly to Google", and calls the Shopping Graph "a dynamic repository of product info". The second is product information that retailers and brands publish on their own websites, including the Product and Offer structured data that Google's Search Central documentation has accepted for merchant listings since September 2022. The third is wider web material: images, descriptions, reviews and YouTube videos. The fourth is local inventory, the stock held in physical shops.

The graph then reconciles records, so that a pair of trainers listed by 40 retailers is understood as one product with 40 offers. It also infers attributes that no merchant typed in. In a February 7, 2023 post, Randy Rockinson, a group product manager for Google Shopping, said the graph uses machine learning to work out traits such as whether a jacket packs down, how it performs in cold weather or whether its materials are sustainable. That inferred layer is what allows a query like "cozy sweaters for happy hour in warm autumn colors" to return products whose feeds never contained those words.

Freshness is the second defining claim. Google said at its agentic checkout launch on November 13, 2025 that about 2 billion listings are updated every hour. It has not said how those updates are distributed, or how quickly a single merchant's price change reaches every surface.

Where it shows up

The Shopping Graph is infrastructure, not a product a shopper opens. It sits behind the Shopping tab and shopping.google.com, product panels in standard Search results, Google Images, YouTube shopping features, price tracking and deal alerts. Google Lens draws on it to identify products in photographs; when Google put Shopping ads into Lens results in October 2024, it cited nearly 20 billion visual searches a month, about a fifth of them shopping-related.

It also grounds Google's generative shopping features. The Gemini-based Shopping redesign rolled out in the US on October 15, 2024 added AI-written product briefs, dynamic filters and a personalised feed, all built on the graph. Shopping inside AI Mode, Google's conversational search interface, and inside the Gemini app uses the same data. The November 2025 launch made Gemini's shopping answers available to all US users with "the same Shopping Graph data as Search".

On the transaction side, agentic checkout lets a shopper set a size, colour and budget on a tracked product; when the price falls within range, Google can complete the purchase on an eligible merchant's site through Google Pay, after the shopper confirms. Wayfair, Chewy, Quince and some Shopify sellers were the first merchants. The graph supplies the product and price data; payment and order handover run on separate systems.

Origin and evolution

The name borrows from the Knowledge Graph, which Google introduced on May 16, 2012 with more than 500 million objects and 3.5 billion facts about the relationships between them, according to a post by Amit Singhal, then senior vice president of engineering. Rockinson's 2023 post says the Shopping Graph was modelled on it.

The 2021 launch came a year after Google made it free for merchants to appear in Shopping results. Ready's announcement said that change had produced a 70% increase in catalogue size and an 80% increase in merchants. It also expanded a partnership with Shopify, allowing its 1.7 million merchants to list products across Google in a few clicks.

Google's own scale figures have climbed with each announcement, and they count listings rather than distinct products:

Date Figure cited by Google
May 2021 more than 24 billion listings
September 2022 more than 35 billion listings
October 2024 more than 45 billion listings
November 2025 more than 50 billion listings
May 2026 more than 60 billion listings

Google has not published how a listing is defined, how duplicates are treated or how the figures are audited. All of them are self-reported.

Why it matters to marketers

For a retailer, the Shopping Graph decides whether a product can be found at all on Google's shopping surfaces, paid or unpaid. Free listings, Shopping ads and Performance Max shopping placements all depend on Merchant Center data that enters the graph, and the conversational features depend on attributes the graph either receives or infers. Google's AI performance insights report, added to Merchant Center on May 27, 2026, measures share of voice and attribute completeness across AI Mode, AI Overviews and Gemini, which treats feed quality as a visibility metric in its own right.

The graph has also become Google's main argument in agentic commerce. Heiko Hotz, a Google AI strategist, wrote in September 2026 that "agentic commerce runs on grounding data!", presenting the 60 billion figure as raw material for shopping agents. Rival assistants building product search from scratch have nothing comparable.

Limitations and disputes

The data is only as current as its inputs. Google says Merchant Center's automatic import refreshes every 24 hours, yet a Google Ads specialist, Emmanuel Flossie, found products nine to ten days out of date in January 2026. A Merchant Center beta spotted in May 2026 builds listings from a one-time AI scan of a website, with no ongoing sync.

Size does not guarantee exposure. Productrise, a vendor that sells product-visibility monitoring, ran more than 100,000 identical queries in July 2026 and found AI Mode showed about 95% fewer product listings than standard search, with products on about 23% of AI Mode queries against 88% in standard results. An August study by the same company found that only 1.28% of products ranking in standard search also appeared in AI Mode, and that the AI Mode lead price for identical products averaged 21.6% higher. Google had not responded publicly to either study, and both carry the methodological limits of a vendor's sample.

Control is the deeper concern. Lindsay Owens, executive director of Groundwork, called Google's commerce plans "an NSA for capitalism" in January 2026, warning that chat and search data could be used for personalised pricing; Google called the claims inaccurate and said merchants cannot show higher prices on Google than on their own sites. In Europe, the history of Google Shopping weighs on any data advantage. A Stockholm court ruled on July 1, 2026 that Google's self-preferencing of its comparison shopping service continued after 2017, awarding Klarna damages including 950 million pounds. The ruling did not concern the Shopping Graph directly, but it shows how a dominant product-data operation attracts competition scrutiny.

The scale numbers also invite caution: PPC Land's October 2024 Lens coverage described the 45 billion figure as products, while Google's own posts say listings.

Not the same as

  • Knowledge Graph: Google's general entity database of people, places and things. The Shopping Graph applies the same idea to commerce, with prices, offers and stock that change hourly.
  • Product feed: The file a merchant submits. A feed is one input to the graph; the graph combines feeds with crawled data, reviews and inferred attributes.
  • Google Merchant Center: The account and tool through which merchants send data and manage eligibility. It is the main entry point, not the data set itself.
  • Amazon's product catalogue: Amazon's records are organised around the ASIN, a page-level identifier that sellers attach offers to inside Amazon's own store. The Shopping Graph indexes offers from many independent shops and sends shoppers, or agents, to them.

Recent developments

Google launched the Universal Commerce Protocol (UCP) on January 11, 2026, an open standard co-developed with Shopify, Etsy, Wayfair, Target and Walmart that lets agents complete purchases without a custom integration for each merchant. At I/O on May 19, 2026, Google cited more than 60 billion listings and announced Universal Cart, a cross-merchant cart that tracks price drops and stock; Vidhya Srinivasan repeated the figure at Google Marketing Live on May 20. In April 2026, Google began placing sponsored tiles inside the Shopping tab's free listing grid. From September 18, 2026, Google enabled native checkout in AI Mode and Gemini by default for matched Shopify stores, notifying merchants by email afterwards. As of October 2026, the graph is the product layer under each of these features, and Google has still published no independent measure of its size or accuracy.

Timeline

  • May 16, 2012 - Google introduces the Knowledge Graph with more than 500 million objects.
  • April 21, 2020 - Google announces free product listings in the US Shopping tab.
  • May 18, 2021 - Google announces the Shopping Graph at I/O, citing more than 24 billion listings, and expands its Shopify partnership.
  • September 2022 - Google Search Central expands merchant listing eligibility through Product structured data.
  • September 28, 2022 - Google says the graph covers more than 35 billion listings.
  • February 7, 2023 - Google publishes an explainer describing the graph as modelled on the Knowledge Graph.
  • October 3, 2024 - Shopping ads come to Google Lens; Google cites more than 45 billion.
  • October 15, 2024 - Gemini-based Google Shopping redesign rolls out in the US.
  • November 13, 2025 - Agentic checkout and Gemini app shopping launch in the US; Google cites more than 50 billion listings and about 2 billion hourly updates.
  • January 11, 2026 - Universal Commerce Protocol and Business Agent launch.
  • May 19-20, 2026 - Google cites more than 60 billion listings at I/O and Google Marketing Live; Universal Cart announced.
  • May 27, 2026 - Merchant Center adds AI performance insights.
  • July 1, 2026 - Stockholm Patent and Market Court rules against Google in Klarna's comparison shopping claim.
  • September 18, 2026 - Native checkout in AI Mode and Gemini enabled by default for matched Shopify stores.

Summary

  • Who: Google builds and operates the Shopping Graph; merchants, brands and Shopify sellers supply much of its data through Merchant Center and their websites; shoppers and, increasingly, AI agents consume it.
  • What: A machine-learning-built data set of products, offers, sellers, brands, reviews and inventory, which Google says holds more than 60 billion listings with about 2 billion updated hourly.
  • When: Announced on May 18, 2021 with more than 24 billion listings, and expanded through the Gemini Shopping redesign in 2024, agentic checkout in 2025 and Universal Cart in 2026.
  • Where: Behind Google Shopping, Search, Images, Lens, YouTube, AI Mode and the Gemini app, with agentic features launched first in the US.
  • Why: Answering shopping queries and completing purchases requires a structured, current map of products and who sells them, and control of that map is central to Google's position in both search advertising and agentic commerce.