AI RACE— The AI Race
Business

Amazon Reportedly Developing 'World Product Graph' to Index Global E-Commerce Data for AI Agents

Amazon is planning an ambitious project to catalog every product sold across the internet, aiming to keep shoppers on its platform and position itself as the default data source for AI shopping agents.

11/14/2026, 06:49
Amazon rò rỉ kế hoạch phát triển 'World Product Graph': tham vọng số hóa toàn bộ dữ liệu thương mại điện tử cho Agent AI

Amazon is working on a sweeping initiative to catalog every product sold on the internet—even those it does not sell itself—according to an internal document reported by Reuters. Known internally as the "World Product Graph," the project aims to turn Amazon into a comprehensive product directory to retain shoppers and capture traffic from emerging artificial intelligence agents.

Keeping Shoppers and AI Bots on Amazon

The project is designed to prevent consumers from leaving Amazon to conduct product research on platforms like Reddit or Google. By providing details on virtually any item available online, Amazon hopes users will land on its pages even when searching for items carried exclusively by external retailers.

The initiative also targets the next generation of online shopping driven by automated software. If Amazon establishes the largest, most complete catalog of global product information, autonomous AI agents are more likely to rely on its database to fulfill user shopping requests. This would ensure that consumers browsing through AI bots continue to direct traffic and transactions through Amazon's ecosystem.

Amazon has already laid preliminary groundwork for incorporating external inventory, introducing a "shop direct" feature last year that surfaced links to products sold by third-party retailers.

Improving Accuracy and Lowering AI Costs

Beyond external search dominance, the World Product Graph is intended to resolve key performance and cost issues within Amazon’s own AI shopping tools. The company currently operates an AI shopping assistant called Alexa for Shopping, which was previously known as Rufus.

According to the internal document, Amazon noted that "many Rufus answers are factually wrong or incomplete," while running user searches through the AI system remains expensive. By feeding the AI a vastly expanded, structured dataset of products, Amazon expects to improve the factual accuracy of the assistant's recommendations while enabling the system to run on smaller, less expensive models.

Roadblocks and Timeline

Bringing the World Product Graph to market faces notable technical and operational hurdles. The internal document reportedly noted that execution "depends on discovering how brands organize their products on their own websites," which could require Amazon to crawl and parse competitors' digital storefronts.

While the document listed an internal target release date of October 1, 2027, the future of the project remains uncertain. In a statement to Reuters, Amazon downplayed the leak: "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."

◗ Sources

The Verge11/14

Related stories