Amazon's World Product Graph plan takes aim at AI shopping agents

An internal Amazon document seen by Reuters describes indexing nearly every product sold online, partly so shopping agents keep relying on Amazon's catalog.

Illustration of a vast glowing web of connected product cards, from sneakers to soda cans, converging on a central warehouse while small robot shopping agents travel along the lines. Image: AgenticFeed illustration

Amazon has drafted a plan to collect product data on nearly every item sold online, including goods it does not stock, in part to keep AI shopping agents relying on its catalog, Reuters reported on Nov. 13. The initiative, called World Product Graph, is described in an internal document reviewed by the news agency and by people familiar with the matter. Amazon has not confirmed the plan and described the document as a brainstorming draft.

The document is a PR/FAQ, the mock press release and question list that Amazon teams write to pitch new products. According to Reuters, it carries an aspirational public release date of Oct. 1, 2027, and the news agency could not determine if or when the plan will be carried out. People familiar with the project told Reuters it could be modified or dropped for financial or other reasons.

"Draft, outdated documents often paint an incomplete picture and the existence of a brainstorming document such as this one doesn't necessarily reflect our current plans," Amazon said in a statement to Reuters.

How the data would be gathered

According to the document, World Product Graph "depends on discovering how brands organize their products on their own websites." Reuters reported that this would involve some brands and retailers sending product data to Amazon on a continuing basis, with Amazon also crawling other retailers' websites on its own. Reuters said it could not determine how the crawling would work or whether Amazon would seek permission in every case.

The document says Coca-Cola and Bayer are among the companies already signed up to contribute data, and that Amazon held a "CEO-level P&G/Amazon summit" for a trial with Procter & Gamble. Those companies did not respond to Reuters' requests for comment. The project sits in Amazon's Selection & Catalog Systems unit under vice president Robert Tekiela, according to the document.

Rufus accuracy and cost

The document also points to problems with Rufus, the AI shopping assistant Amazon began calling Alexa for Shopping in May. "Many Rufus answers are factually wrong or incomplete, but customers lack expertise to detect these errors," it states. It puts usage at about 83 million searches a week, calls running them "expensive" and says more complete data would let Amazon use cheaper models more efficiently.

Amazon told Reuters the usage figures are "significantly outdated" and that its own data shows Alexa for Shopping is roughly 97% accurate globally. CEO Andy Jassy has said the assistant was used by more than 400 million people in the past year, Reuters reported.

The agent angle

People familiar with the matter told Reuters that one goal is to hold off outside AI shopping agents, which Amazon executives worry could steer customers to sites with fuller product data. With the largest store of product information, Amazon could make agents acting for shoppers look for products on Amazon.com rather than on other retailers' sites, those people said. That would also protect Amazon's growing advertising business, according to the report.

On Amazon's third-quarter earnings call, Jassy said "few customers are doing much shopping this way yet" and that the advertising business had seen minimal impact, Reuters reported.

The plan surfaces as Amazon limits outside agents on its store. Amazon blocked Meta's Muse agent from its retail site in September and is in an ongoing lawsuit against Perplexity over similar access, according to Reuters, which said Muse has drawn more than 10 million users in the US and Canada since it launched in September.

Reuters reported on a 2025 Morgan Stanley research report about where shopping starts. It found that 16% of online shoppers begin product research on Amazon, compared with 57% on Google. When shoppers already know which product they want, Amazon leads, with 41% of those searches starting there against 23% for Google, the report found.

Employees have described World Product Graph as an early step toward making Amazon a place to buy almost anything, according to Reuters. Reuters reported that it would extend the existing Buy for Me and Shop Direct services, which list products sold on other sites and let customers buy them with Amazon acting as the intermediary.

Source: Reuters (via CNA), “Exclusive: Amazon's plan to own the world's product data and fight AI shopping agents”

Additional sources: The Verge, “Amazon has a plan to index the world's online shopping data”

Get the AgenticFeed newsletter

A free, regular roundup of the agent stories that matter. No spam, and you can unsubscribe at any time.

Want to reach readers following AI agents? Advertise with AgenticFeed

Next event · 19 Nov, 2026

MCP Dev Summit Nairobi

The Linux Foundation’s Nov. 19 and 20 summit brings Model Context Protocol developers to Kenya, with keynotes from OpenAI, GitHub and the Gates Foundation.

The AgenticFeed newsletter

A free, regular roundup of the agent stories that matter. No spam.