In short
The Shopping Graph is Google's large product database built with machine learning. You can think of it as the product counterpart to the better-known Knowledge Graph: while the Knowledge Graph links entities such as people, places and organisations, the Shopping Graph arranges products, brands, retailers, prices, availability, attributes and reviews into a coherent picture.
Google describes it as a data set with billions of product listings that helps the system understand what a product is, who sells it and how it relates to similar products. At Google's developer conference, its size was put at over 50 to 60 billion product listings and billions of updates per hour.
The Shopping Graph is so important for online retailers because Google's AI shopping draws precisely from it. Both AI Overviews and AI Mode rely on the Shopping Graph for product information. Anyone not represented there simply cannot appear in these results – regardless of how well their own website ranks in classic search. The Shopping Graph is therefore less an optimisation detail than a bouncer: it decides who is even in the running in Google's AI-supported shopping experience.
How the Shopping Graph is fed
The graph draws its data from several sources and merges them into a normalised product entity. The two most important channels for retailers are:
- The product feed in Google Merchant Center. This is the canonical, primary data source. Via the feed, a retailer supplies title, description, price, availability, GTIN, brand and other attributes in a structured way. Without an up-to-date Merchant Center feed, products do not get into the Shopping Graph in the first place.
- The schema.org
Productmarkup on the shop pages. Structured data directly on the product page supplements the feed and provides additional signals that can be verified on the page.
In addition, Google enriches the graph with information it crawls on the web: product descriptions, images, user reviews and signals from third-party sites. Machine learning processes all of this in order to recognise not only that a product exists but also its properties – material, colour, size, intended use – and how it relates to alternatives. If two retailers describe the same product differently, the Shopping Graph normalises this information into a single, standardised entity.
Feed and markup are separate systems
A common misconception is that the Merchant Center feed and page-based markup are interchangeable. They are not. They are two separate input channels that both flow into the Shopping Graph but play different roles. The feed is the binding data stream controlled by the retailer; the markup is the confirmation of the same facts visible on the page itself. For maximum coverage, both should be present, complete and, above all, free of contradictions – price information in the feed and on the page must not diverge.
Why the Shopping Graph matters for Shopware shops
For a Shopware operator, the consequence can be put bluntly and simply: without an up-to-date Merchant Center feed, the products do not appear in Google's AI shopping. Data quality in the feed is the primary signal here, not some keyword trick. The practical task is to keep the product feed clean, complete and current – ideally regenerated automatically after every catalogue change.
For many mid-sized companies, this is also reassuring news. Anyone who already maintains Merchant Center and product data for Google Shopping has already done a large part of the Google-side homework for AI visibility. The remaining work then consists more of closing gaps and adding data depth than of a completely new project.
A concrete example
A Shopware shop for outdoor equipment maintains a complete Merchant Center feed with GTINs, correct prices, stock levels and shipping details. If a user asks Google's AI Mode for a "waterproof hiking jacket under 200 euros with good breathability", Google can pull this shop's products from the Shopping Graph, match them against the query and suggest them in the AI answer – including price and availability. A competing shop without a maintained feed remains invisible in this answer, even if it had the more suitable product. Visibility is decided by the data pipeline, not by the marketing budget.
Shopping Graph, SEO and agentic commerce
With the Shopping Graph, Google's product ecosystem is merging: organic search, Merchant Center, structured data, reviews, prices and stock levels flow into the same data basis. The consequence is that performance increasingly depends on the accuracy of the entire product data pipeline and no longer just on keywords or backlinks. Precisely this clean, machine-readable data foundation is also the basis for agentic commerce – i.e. for AI agents that not only recommend but buy on the user's behalf. Anyone who puts their data in order for the Shopping Graph is thus also building the prerequisite for the next stage of AI-supported shopping. Further details on structure and how it works are provided by the product data specification in the Google Merchant Center Help.
The Shopping Graph and the change in product search
The Shopping Graph is more than a technical detail in the background – it stands for a fundamental shift in how people find products. Traditionally, product search began with a typed query and ended in a list of blue links and ads. In the AI-supported model, the user formulates a complete need in natural language – for example "I'm looking for a quiet, energy-efficient model for a small flat with delivery by Friday" – and expects a curated answer instead of a list of results. For Google to be able to build such an answer, it must not only know products but also understand their properties, prices and availability and weigh them against each other. That is exactly what the Shopping Graph does.
For retailers, this shifts the lever. It is no longer enough to optimise for a single keyword. What becomes decisive is how complete, correct and attribute-rich the product data is, because the graph builds from it the picture against which every natural-language query is matched. A product with maintained material, a precise size specification, the correct energy class and clean shipping information simply has more points of contact in this model at which it can match a specific query.
Which data fields count most in the Shopping Graph
Not all feed attributes are equally important. Some form the foundation for inclusion, others decide how well a product fits specific queries. A rough prioritisation:
| Field | Function in the graph |
|---|---|
| GTIN / MPN | unique identification and merging of several retailers into one product entity |
| Title & description | semantic classification, matching with natural-language queries |
| Price & availability | filtering by budget and deliverability, often an exclusion criterion |
| Brand | link to the brand entity, trust signal |
| product-specific attributes (colour, size, material, energy class) | accuracy of matches for detailed queries |
| Shipping & returns | increasingly relevant for transactional AI answers |
Incomplete attributes are the most common reason why a product is in the graph but is not shown for specific queries. Anyone who only maintains title and price is visible for "coffee machine" but invisible for "bean-to-cup coffee machine with milk system under 600 euros, available immediately" – and that is exactly where the purchase intent lies.
The role of data quality
The Shopping Graph rewards consistency and punishes contradictions. If the price in the feed differs from the price on the page, or if the feed reports availability while the product page shows "sold out", this undermines trust in the data source – in the worst case, the products concerned are held back. For a Shopware shop, this specifically means: the feed export should be coupled directly to the leading data source, not to a manually maintained interim state that quickly becomes outdated. Automation is not a convenience here but the prerequisite for reliable visibility.
Distinction: Shopping Graph, Knowledge Graph and feed
Because many similar-sounding terms circulate around product search, a clean distinction is worthwhile. The Knowledge Graph is Google's general knowledge database about entities in the world – people, places, organisations, concepts. The Shopping Graph is the specialised counterpart for products: the same basic idea of a networked entity database, but tailored to commercial objects and enriched with prices, availability and retailer references. The product feed, in turn, is not a database but an input channel – the structured data stream with which a retailer feeds its products into the Shopping Graph.
This distinction is more than academic. It makes clear where a retailer actually has influence: not on the graph itself, but on what it puts into it. The quality of the feed and the page markup is the only directly controllable lever. Everything else – normalisation, linking with reviews, classification relative to competing products – happens within Google's system and is beyond direct control.
What retailers can do in concrete terms
From the way the Shopping Graph works, clear steps can be derived that apply regardless of shop size:
- Set up the Merchant Center account completely and couple the product feed to the leading data source, so that prices and stock levels stay up to date automatically.
- Maintain all available identifiers – in particular GTIN and MPN – so that products are clearly assigned to an entity.
- Fill in attributes beyond the required fields: material, colour, size, energy class, intended use. Every additional, correct attribute increases the chance of matching a specific query.
- Keep feed and page markup free of contradictions. Discrepancies between the two undermine trust in the data source.
- Provide shipping and returns details, because these are increasingly taken into account for transactional AI answers.
The common denominator of these measures is unspectacular: it is about data quality and timeliness, not a trick. That is exactly what makes the Shopping Graph a fair chance for well-organised retailers – visibility follows the data work, not the budget.
Frequently asked questions about the Shopping Graph
Is the Shopping Graph the same as Google Shopping? No. Google Shopping is the visible surface with product ads and listings. The Shopping Graph is the product database behind it, from which these surfaces – and now also AI Overviews and AI Mode – draw.
Can I get into the Shopping Graph without Merchant Center? Not reliably in practice. The Merchant Center feed is the canonical source. Page markup alone is a supplement, not a full substitute for inclusion.
How up to date does my data need to be? As up to date as possible. The graph is updated continuously, and outdated prices or stock levels can lead to products being displayed incorrectly or filtered out. An automated feed export after every catalogue change is the robust way.
Does the Shopping Graph also help with ChatGPT or Perplexity? Not directly – it is a Google system. However, the underlying work, i.e. clean structured product data, also benefits visibility in other AI systems, because they too prefer machine-readable facts.