Amazon’s World Product Graph: whoever maps every product needs your data

Published on November 16, 2026

According to an internal document reported by Reuters, Amazon is planning a database for virtually every product on Earth, including products Amazon does not sell. What is known about the World Product Graph and why manufacturers should secure control over their own product data now.

Diagram: a product in the World Product Graph. On the left, one verified version from PIM and DAM and contradictory data crawled from retailers; on the right, search, shopping agents and marketplace reading the graph
“The map is not the territory.”
— Alfred Korzybski

On November 13, 2026, Reuters reported on an internal Amazon document describing a project called the World Product Graph. The goal: data on virtually every product for sale anywhere in the world. Including products Amazon does not even carry.

Until now, the roles were clear. Manufacturers maintain their product data, retailers and marketplaces take it over, and every channel builds its own version from it. The World Product Graph turns this around. One platform wants to draw the single map of all products that buyers, and increasingly AI shopping agents, will navigate by.

That raises a new question for every manufacturer and retailer.

If Amazon draws the map of every product: whose data is on it?

What is known about the World Product Graph?

The source is an internal document in Amazon’s typical PR/FAQ format, a kind of press release written in advance, with questions and answers. Reuters reviewed it and checked it with people familiar with the plan. According to the report:

  • Scope: The graph is meant to cover data on practically every item sold worldwide, whether or not Amazon stocks it.
  • Data sources: Some brands and retailers are to supply product data continuously themselves. Other retailers’ websites Amazon plans to crawl on its own. Reuters could not establish how exactly, or whether permission would be needed in every case.
  • Partners: According to the document, Coca-Cola and Bayer have already agreed to contribute product data. There was also a CEO-level meeting with Procter & Gamble about a trial.
  • Ownership: The project sits in the Selection & Catalog Systems unit under Vice President Robert Tekiela.
  • Timeline: The document names October 1, 2027 as its aspirational date. According to one source, Amazon is already reassigning staff internally to the project.

Amazon itself plays it down. The document is an outdated draft from a brainstorming session, it says, and does not necessarily reflect current plans. At the same time, the company says it is constantly working to make its catalog comprehensive, accurate and helpful. Reuters also notes that the project could still be changed or abandoned.

So whether the World Product Graph arrives exactly like this is open. The direction is not.

Why is Amazon building a map of every product?

According to Reuters, Amazon executives are increasingly wary of AI shopping agents steering buyers toward sites with more complete product data. Meta’s agent Muse has gained more than ten million users in the US and Canada since its launch in September. Amazon blocked Muse from its website in the same month and has been in a legal dispute with Perplexity since last year to prevent similar access.

Then there is a figure Reuters cites from a 2025 Morgan Stanley study. Only 16 percent of online shoppers start their product research on Amazon, compared with 57 percent on Google. Only when shoppers already know exactly what they want does Amazon lead, at 41 percent.

The internal document names the problem openly: Amazon cannot be the destination for every shopping mission as long as customers leave for Google, brand websites or review sites to do their research. Amazon’s own shopping assistant, formerly Rufus and now Alexa for Shopping, also gives many incomplete or wrong answers, according to the document. Amazon disagrees, citing an accuracy of roughly 97 percent, and says the figures in the document are significantly outdated.

The logic behind it is clear all the same. Whoever hands purchase decisions to agents needs a product data foundation those agents can rely on. Whoever owns that foundation owns the place where people search, compare and buy.

What does this mean for manufacturers and retailers?

One sentence in the document is particularly revealing for manufacturers. The World Product Graph depends on discovering how brands organize their products on their own websites.

In other words: the map is made from your data. Either you supply it yourself, or Amazon assembles it from whatever it finds on your website and at your retailers.

An example. A manufacturer offers a cordless vacuum cleaner in two versions, X2 and X2 Pro, each in three colors. Its own website lists the weight in grams, retailer A in kilograms, and retailer B not at all. At one retailer, the Pro version is still linked to the previous model, and one of the images shows the wrong color. These are exactly the questions the graph is meant to settle, according to the document: which is the newest model and which color variant belongs to which product. If the manufacturer does not supply one clear version, an algorithm decides. And when in doubt, a shopping agent recommends the competing product whose data is free of contradictions.

This leads to a principle I call source authority: whoever supplies the most complete and consistent version of a product determines how that product appears in catalogs, search engines and agents.

The question is no longer whether your product is in the graph. The question is whose version.

Is this only about Amazon?

No. The World Product Graph is the most ambitious version of a development that is already well underway. Every AI search and every shopping agent builds its own picture of your products, from feeds, websites, marketplaces and reviews. With open protocols for agentic commerce, this picture becomes the basis for actual purchases.

For manufacturers and retailers, this means it is no longer enough to maintain product data channel by channel. What is needed is one well-maintained truth about every product, published in every direction. To Amazon, to Google, to your own website and to every agent that asks for it.

What does this mean for PIM and DAM?

The PIM turns from an internal maintenance tool into the authoritative source for the outside world. The DAM is part of it, because images, color variants and documents are part of the product truth, not an add-on. These points decide whether a graph understands your products correctly:

  • Unique identifiers: GTIN, manufacturer part number and model name, consistent across all channels.
  • Variants and successors: Which color, which size, which model replaces which. This is exactly what Amazon wants to settle, according to the document.
  • Complete attributes: Technical data, dimensions, weight, material and regulatory information in consistent units.
  • Correctly assigned media: Every image belongs to the right variant, with alt text and metadata.
  • Structured data on your own website: If graphs learn how brands organize their products, your own website is a data source and not just a shop window.
  • Origin and freshness: Where does a value come from, and when was it last checked?
  • Every language from one record: One product, many markets, no copies drifting apart.

How ainavio secures your source authority

ainavio is a PIM with AI and an integrated DAM. It is built to turn scattered product data into one reliable version and bring it into every channel.

  • In the Golden Record, ainavio merges product data from several sources into one central record per product. Sources can be prioritized so that the most reliable value wins.
  • The Research Agent researches missing manufacturer and product information and shows the source and a confidence score for every attribute.
  • The Quality Agent finds missing information, duplicates and contradictions before a channel or an agent does.
  • In the DAM, media are linked to the right products automatically. The Media Agent adds alt texts, image descriptions and metadata.
  • The Marketplace Agent maps categories and mandatory attributes to the fields of 26 shops and marketplaces, Amazon among them. Before every export, ainavio checks completeness and channel compliance.
  • The Translation Agent translates product data into any language and converts units of measurement for each market.

Whatever AI adds lands as a suggestion in the PIM and is approved by your team. That way, control over your product data stays with you, whether Amazon, Google or an agent is asking. How ainavio connects to AI assistants through MCP is described in our post on the orchestration layer.

Back to the map

Korzybski’s line comes from the philosophy of language, but it fits the commerce of tomorrow surprisingly well. Amazon wants to draw the map of every product. The territory remains your products and the data you maintain about them.

A map is only ever as accurate as the survey behind it. Whoever keeps their product data complete and consistent decides what appears on the map. Whoever does not will be surveyed by others.

If you do not maintain your product data yourself, others will describe your products for you.

Frequently asked questions

What is the Amazon World Product Graph?

An internal Amazon project reported by Reuters in November 2026. It is meant to bring together product data on virtually every item sold worldwide, including products Amazon does not sell itself.

Has the World Product Graph been officially confirmed?

No. Amazon describes the document as an outdated draft from a brainstorming session. Reuters could not establish whether or when the project will be implemented.

When is the World Product Graph supposed to launch?

The internal document names October 1, 2027 as its aspirational date. According to Reuters, it is unclear whether that date is realistic.

Do manufacturers have to supply their product data to Amazon?

Nothing is known about that. According to the document, some brands contribute voluntarily, including Coca-Cola and Bayer. For other products, Amazon plans to analyze retailer websites. Manufacturers who supply data themselves, or structure their own data cleanly, have more influence over how their products appear.

What should manufacturers and retailers do now?

Build one central, complete and consistent version of every product, with unique identifiers, clean variant logic, correctly assigned media and structured data on their own website. A PIM with an integrated DAM is the foundation for this.

Your project
What would this look like for your product data?

Talk to us about your range, your channels and your legacy systems. We will show you where PIM with AI fits into your setup.

Björn Thomsen, Head of Marketing at ainavio

Björn Thomsen

Head of Marketing, ainavio

Björn Thomsen is Head of Marketing at ainavio, the PIM with AI. His topic is how e-commerce product data works for both people and machines, from the online shop to AI search.

contact@ainavio.com
+49 (0) 2842 - 929987-3

AI in Product Data Management

From raw material to enriched content: The art and science of AI-based product onboarding

Weiterlesen
AI in Product Data Management

Behind the scenes of the AI workflow: How multiple LLMs are working together to deliver future-oriented product data

Weiterlesen
AI in Product Data Management

PIM and AI: Why artificial intelligence is essential in your product data strategy

Weiterlesen

Fangen Sie noch heute an!

Ihre Strategie. Ihr Tempo. Ihr Wachstum. Unsere Unterstützung – auf ganzer Linie.