Amazon plans World Product Graph to index every online product

Amazon is working on a plan to collect data on virtually every product sold anywhere on the internet, including products it does not sell, according to Reuters, as relayed by The Verge on November 13, 2026. The project is called the “World Product Graph.” Reuters bases its report on an internal Amazon document and on people familiar with the matter.

A product graph is a database that stores not only product listings but also the relationships between them: which model replaces which, which accessory fits which device, which option suits which need. An AI agent, in this context, is software that carries out shopping tasks on a user’s behalf, such as comparing products or placing an order.

Why does Amazon want to own the world’s product data?

According to Reuters, the goal is to make sure shoppers never need to leave Amazon to research a purchase. The Verge writes that the project is meant to make it less likely that shoppers go to Reddit or Google for product research. The internal document says Amazon “cannot be the go-destination for every shopping mission if customers must leave to research product relationships on Google, brand websites, and review sites,” as quoted by Reuters.

The figures cited by Reuters show the gap. According to a 2025 Morgan Stanley research report, 16% of online shoppers start their product research on Amazon, against 57% on Google. When customers already know which product they want, Amazon leads: 41% of searches begin there, compared with 23% on Google and 18% on Walmart, according to the same report.

The same logic applies to AI agents. According to The Verge, if Amazon holds the largest stockpile of shopping data, agents may rely on its index rather than on others, so even people shopping with bots would end up on Amazon pages. Reuters adds, citing people familiar with the matter, that this would also help protect Amazon’s growing advertising business.

What problem with Alexa for Shopping would it fix?

Amazon’s AI shopping assistant, known as Rufus until May, is now called Alexa for Shopping, according to Reuters. The internal document states that “many Rufus answers are factually wrong or incomplete, but customers lack expertise to detect these errors,” according to Reuters. It also says that users run about 83 million searches per week with the assistant and that running them is “expensive.” More data would let Amazon use cheaper models more efficiently, the document argues, according to Reuters and The Verge.

Amazon disputes these points. In a statement to Reuters, the company said the usage figures in the document are “significantly outdated” and that, according to its own data, Alexa for Shopping is roughly 97% accurate globally. CEO Andy Jassy said the assistant had been used by more than 400 million people over the past year, according to Reuters.

How would Amazon collect the data?

The document says the project “depends on discovering how brands organize their products on their own websites,” as reported by Reuters and The Verge. According to Reuters, that would involve some brands and retailers sending product data to Amazon continuously, but also Amazon crawling other retailers’ websites on its own. Crawling means automatically visiting web pages to read and copy their content. Reuters could not determine how Amazon would crawl those sites or whether it would need permission in every case.

On the brand side, the document says Amazon has already signed up Coca-Cola and Bayer, among others, to contribute product data, according to Reuters. It also mentions a “CEO-level P&G/Amazon summit” for a trial with Procter & Gamble. According to Reuters, the companies did not respond to requests for comment.

How do rival AI shopping agents factor in?

According to Reuters, Amazon executives have grown wary of agents that could send customers to sites with more complete data. Jassy tried to downplay the threat during the third-quarter earnings call, saying that “few customers are doing much shopping this way yet” and that the advertising business has been minimally affected, as quoted by Reuters. Reuters reports that Meta’s Muse agent has passed 10 million users in the United States and Canada since its September launch, and that Amazon blocked it from its retail site the same month. Amazon also sued the AI company Perplexity last year to prevent similar access, a case that Reuters says is still ongoing.

According to Reuters, employees describe the World Product Graph as the start of an effort to turn Amazon into a one-stop shop, extending the existing “Buy for Me” and “Shop Direct” services. The Verge notes that Shop Direct, launched last year, already shows links to outside retailers’ products on Amazon.

Will it actually ship?

Nothing is certain. The document is a PR/FAQ, which Reuters describes as a press release and frequently asked questions, typically a six-page narrative memo laying out the rationale for a new product. It carries an aspirational release date of October 1, 2027, but Reuters could not determine whether the plan will be implemented or when. The people interviewed said it could be modified or aborted for financial or other reasons. According to Reuters, the project belongs to the Selection & Catalog Systems unit and is organized under vice president Robert Tekiela.

Amazon’s statement to Reuters stresses that “draft, outdated documents often paint an incomplete picture” and that a brainstorming document does not necessarily reflect its current plans.

What it changes for a US company

For a US brand or retailer, the main stake is control of product data. If Amazon builds a graph from brand websites, the way a brand structures its product pages, specifications and variants could shape how agents and shoppers see it on Amazon. Companies may want to audit how complete and consistent their product data is, decide whether to contribute it directly, as Coca-Cola and Bayer reportedly have, and check what their site rules say about crawlers. Retailers that compete with Amazon should also watch whether a single index becomes the default source for shopping agents. These are cautions, not certainties: the plan remains a draft.