Amazon Wants to Build a Universal Product Graph That Includes Items It Doesn't Even Sell

Amazon is plotting a project called "World Product Graph". According to Reuters, the company wants to build a massive data vault covering every item for sale anywhere on the internet: even products not sold on Amazon's own marketplace will have their information collected, and delivered directly to shoppers through Amazon's website.

This sounds like a general product encyclopedia at first, but its actual core purpose is to seize the shopping search entry point. The Verge's report notes that Amazon hopes to reduce how often shoppers leave its site to do pre-purchase research on Reddit, Google and other platforms; it also wants users to end up on Amazon even when they were originally searching for products from other sellers. The same logic applies to AI shopping agents: whichever party holds the most complete product dataset, the more likely the agents are to use that party's index as their default answer.

Reuters described the project formally as the "World Product Graph". It is built to resolve the long-standing fragmentation of product search: product specifications, reviews, pricing and applicable scenario details are scattered across brand sites, forums, price comparison platforms and different e-commerce marketplaces. Amazon already has full data for items sold on its own platform, but it has a critical gap in data for products hosted outside its ecosystem. Once it fills that gap, it will no longer just be a goods retailer — it can position itself as the universal query layer for all product information.

There is a counterintuitive point here: Amazon does not need to sell every product across the internet to profit from cross-platform product search. As long as users open Amazon first when making purchase decisions, the company gets access to advertising revenue, traffic redirection opportunities and control of the transaction entry point. This logic is even more straightforward in the AI agent era: agents do not browse individual web pages, they read structured data directly. Whoever provides the index is far more likely to be called on as a source.

Of course, the plan is still a long way from full rollout. Where to source all the data, how to keep it updated, how to handle data errors and conflicting information, and how to avoid turning third-party product listings into a free traffic entry point for its own site, are all major unresolved issues. Reuters' coverage outlines Amazon's planned project, not a finished live product. For ordinary users, there is no need to change your regular shopping habits in the short term; for professionals working in e-commerce and AI search, it is well worth tracking whether this product graph will become a new layer of foundational internet infrastructure.

Source: The Verge, Reuters.