Amazon is drafting a plan to collect data on virtually every product for sale on Earth — whether it stocks the item or not — according to an internal document reviewed by Reuters. The project, called World Product Graph, is Amazon’s answer to a problem its own memo names plainly: shoppers who cannot get complete product answers on Amazon leave to research elsewhere, and AI shopping agents are learning to take them there. Why it matters: if Amazon owns the world’s product data, the agents doing your shopping may have little choice but to come to Amazon.
November 13, 2026
What the document describes
World Product Graph would assemble deep, structured data on goods far beyond Amazon’s own catalog, so the store can answer questions like which model of a vacuum or tablet is the newest, and which option best fits a shopper’s history. The document says the effort depends on learning how brands organize products on their own websites: some brands and retailers would feed data to Amazon continuously, while Amazon would also crawl other retailers’ sites on its own. Reuters could not determine how Amazon planned to do that crawling, or whether permission would be sought in every case.
The project sits in Amazon’s Selection & Catalog Systems unit, organized under vice president Robert Tekiela. The document is a PR/FAQ — Amazon’s internal format that drafts a future press release to make a product’s case. Its aspirational release date is October 1, 2027. Coca-Cola and Bayer are already signed up to contribute product data, the document says, and Amazon held a CEO-level summit with Procter & Gamble about a trial. Neither company commented to Reuters.
The memo’s blunt admission about Rufus
The document is unusually candid about Amazon’s own AI shopping assistant, Rufus — renamed Alexa for Shopping in May. It says many Rufus answers are factually wrong or incomplete, and that customers lack the expertise to spot the errors. It puts Rufus usage at roughly 83 million searches a week and notes those searches are expensive to run; richer product data would let Amazon answer them with cheaper models. Amazon pushed back on those figures in its statement to Reuters, calling the document’s usage numbers significantly outdated and saying its own data shows Alexa for Shopping is roughly 97% accurate globally.
A defense against AI shopping agents
The second motive, according to people familiar with the matter who spoke to Reuters, is defensive. AI agents that shop on a customer’s behalf could route traffic to whichever retailer holds the best data — and Amazon has already moved against them directly, blocking Meta’s Muse agent from its retail site in September and suing Perplexity last year over similar access, a case still ongoing. Muse has passed 10 million users in the US and Canada since its September launch, and JPMorgan has said it could become the most widely used consumer AI application since ChatGPT. On Amazon’s third-quarter earnings call, CEO Andy Jassy downplayed agents, saying few customers shop that way yet; the memo’s logic is that owning the deepest product data would compel those agents to seek products on Amazon anyway, protecting both traffic and the advertising business built on it.
Why it matters
Read the caveats first, because they are real: this is a brainstorming document, not a launched product. Amazon told Reuters that “draft, outdated documents often paint an incomplete picture” and that the memo does not necessarily reflect its current plans, and Reuters notes that not every PR/FAQ becomes a service — this one could still be changed or dropped. But the direction of travel is hard to miss, and it matches what Amazon is already doing with services like Buy for Me and Shop Direct, which list other retailers’ products for purchase through Amazon. For shoppers, the fight over AI agents is quickly becoming a fight over who owns the underlying product facts those agents read. Whoever holds that data holds the shelf every agent shops from — and Amazon would rather that shelf be its own.
Frequently asked questions
What is Amazon’s World Product Graph?
It is a project described in an internal Amazon document reviewed by Reuters: a plan to gather structured data on nearly every product sold anywhere, whether Amazon stocks it or not, with an aspirational release date of October 1, 2027.
Is World Product Graph confirmed or launched?
No. Amazon calls the document a draft, outdated brainstorming memo that does not necessarily reflect its current plans. Reuters could not determine whether or when the plan will be implemented.
Why would Amazon collect data on products it does not sell?
Per the document, so shoppers can research and compare products without leaving Amazon — and so AI shopping agents, which rely on complete product data, are drawn to Amazon rather than rival retailers.
Which brands are already involved?
The document says Coca-Cola and Bayer have signed up to contribute product data, and that Amazon held a CEO-level summit with Procter & Gamble about a trial. The companies did not comment to Reuters.
Sources: Reuters (internal document review, published November 13, 2026); AI Weekly summary of the Reuters report.