Everything, arranged for the machine to read.
By Ellis Ward | November 14, 2026
Amazon wants to know about every product on earth before an AI agent gets there first.
Reuters reported Thursday that the company is developing an internal project called World Product Graph, an initiative designed to gather data on virtually every product for sale globally. The goal is to make Amazon the authoritative source for product information across the entire retail world, covering inventory far beyond what Amazon itself sells. If the project works, shoppers and AI agents alike would have little reason to look anywhere else.
What Amazon Is Building
The project sits inside Amazon's Selection and Catalog Systems unit and is being organized under Vice President Robert Tekiela. Reuters reviewed an internal planning document written in Amazon's standard PR/FAQ memo format, typically around six pages, that lays out the rationale and architecture for the initiative. Amazon gave the project an aspirational launch date of October 1, 2027. Reuters could not determine whether that date is realistic or whether the project will be implemented at all.
The mechanics depend on brands doing substantial work of their own. Amazon is asking companies to organize product information from their own websites and share it with Amazon's platform. That data would feed into the World Product Graph and, by extension, into Rufus, Amazon's AI-powered search tool. Two companies have already agreed to contribute: Coca-Cola and Bayer. Reuters identified both as early signatories to the arrangement. Neither company publicly confirmed the partnership in the Reuters report, and Reuters did not identify additional brand participants at this stage.
Internally, Amazon is shuffling staff to support the initiative. The document Reuters reviewed does not specify budget or headcount, and Reuters could not independently verify those figures.
The Threat Amazon Is Responding To
This report comes in the middle of a story RFTUV has been tracking: the rapid rise of AI shopping agents and what they mean for retailers who have long depended on owning the customer's first click.
On Thursday, OpenAI launched ChatGPT Wallet, a feature that lets AI models complete purchases on a user's behalf. The pattern across this wave of tools is consistent. Agents do not send users to retailer websites. They route queries through their own systems, complete transactions, and return results without requiring a user to open a browser and navigate to any particular store. That cuts retailers out of the discovery process entirely.
Amazon is not waiting to find out how that plays out. Reuters reported that Amazon blocked a shopping agent called Muse from its retail site this month. The timing is not incidental. Amazon sees third-party AI agents as an active competitive threat, and World Product Graph appears to be the longer-term answer to that threat: it is not just blocking agents at the door, but becoming the data source that any agent would have to draw from.
The logic is durable in a way that individual blocks are not. Amazon can block Muse today. A new agent can appear tomorrow. But if Amazon controls what AI systems know about products at a data layer, it maintains influence over what those systems recommend even when users never visit Amazon.com directly. The World Product Graph is not primarily a retail play. It is a data infrastructure play that resembles one.
A Memo, an Ambiguity, and a Company Non-Denial
Reuters described the source document as a brainstorming memo in Amazon's standard internal format. That format is widely used at Amazon for evaluating new initiatives. Its existence does not confirm that the project is funded, fully staffed, or approved for launch. Internal PR/FAQ documents are written speculatively and frequently, and many never move past the idea stage.
Amazon's response to the Reuters report reflects that ambiguity. "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," the company said in a statement.
That response does not confirm the project is dead. It does not confirm it is moving forward. Amazon uses similar language routinely when confronted with internal plans that reached the press ahead of schedule. The statement is a hedge, not a denial.
The October 1, 2027 date in the document is aspirational, by Reuters' characterization. Reuters could not determine whether that target is achievable. Large-scale data infrastructure projects at a company the size of Amazon typically require years of sustained organizational effort and clear executive priority. Whether World Product Graph has that alignment is not known from the reporting available.
What Remains Unknown
Reuters could not determine how many brands Amazon has approached about contributing data, how many have declined, or what incentives Amazon is offering to participate. Those gaps matter considerably for understanding whether World Product Graph is an early-stage concept or a project with real organizational momentum behind it.
Coca-Cola and Bayer are large, structured organizations with the capacity to do sophisticated data work. Scaling a model that depends on brand participation to cover virtually every product for sale globally requires far more than two anchor partners. The logistics of standardizing product data across thousands of manufacturers, in dozens of formats, across multiple languages and regulatory environments, represent a substantial challenge that the Reuters report does not address in detail.
The reporting does establish that Amazon is thinking systematically about the data layer of commerce, not just the retail layer. The World Product Graph, if it proceeds, would position Amazon as infrastructure for the AI shopping economy rather than merely a participant in it. That is a different ambition than running a marketplace. It is closer to owning the pipes.
Whether Amazon gets there is a separate question. For now, it is a leaked internal document, two confirmed brand participants, an aspirational date, and a company response that neither confirms nor denies the plan.
The AI shopping agent story is not slowing down. Amazon's move to get ahead of it is on the record.
