Amazon is drafting plans for a “World Product Graph”, a database holding details of products sold anywhere on the internet, including ones Amazon does not sell, Reuters reports. The data would be offered through Amazon’s own site.
The aim, as the plan describes it, is to stop shoppers leaving Amazon to research a purchase on Reddit or Google. It is also aimed at AI shopping agents: if Amazon holds the biggest pile of product data, bots comparing prices may lean on its index and send their users to Amazon pages.
The internal document also admits a weakness in Amazon’s own assistant, now called Alexa for Shopping and earlier known as Rufus. It says “Many Rufus answers are factually wrong or incomplete”, and that running shoppers’ searches through the AI is expensive. Feeding it more product data is the proposed fix, and the document says that would also let Amazon run cheaper models.
Amazon has taken a first step with “shop direct”, a feature it introduced last year that puts other retailers’ items in front of its own shoppers. The graph would go much further, and the document says it depends on discovering how brands organise their products on their own websites. That could mean Amazon crawling other retailers’ sites.
Amazon’s statement on the document says that “Draft, outdated documents often paint an incomplete picture and the existence of a brainstorming document such as this one doesn’t necessarily reflect our current plans.”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 document nonetheless carries a release date of 1 October 2027, though it is unclear whether or when the plan will actually be put into practice.
