Amazon Plans Global Product Database to Anchor AI Shopping, Curb Google Search Leakage

Amazon is working on an internal initiative to build a comprehensive database of virtually every product sold worldwide, including items it does not stock, according to a report by Reuters citing an internal document and people familiar with the matter. The project, named World Product Graph, is designed to reduce the need for shoppers to leave Amazon for Google, brand websites, or review platforms when researching purchases.

The draft document reportedly calls for brands and retailers to share product data continuously, and for Amazon to gather information from other retailers' websites. Reuters could not confirm how the company would collect that data or whether it would require permission. The memo indicates that Coca-Cola and Bayer have agreed to contribute data, while Amazon held talks with Procter & Gamble executives about a possible trial. All three companies did not respond to requests for comment.

Such a catalog could help shoppers identify the latest version of products such as tablets or vacuum cleaners, compare models and colors, and receive recommendations based on their shopping history. It would expand on existing features like Shop Direct and Buy for Me, which already allow customers to discover products sold by other retailers through Amazon.

The strategic rationale is clear from search behavior data. A 2025 Morgan Stanley report cited by Reuters found that 57 percent of online shoppers begin product research on Google, compared with 16 percent on Amazon. However, when shoppers already know what they want, 41 percent start on Amazon, ahead of Google at 23 percent and Walmart at 18 percent. A broader catalog could help Amazon reach consumers earlier, while they are still deciding what to buy.

The draft also notes that Amazon's AI shopping assistant, renamed Alexa for Shopping in May after previously being called Rufus, often provided wrong or incomplete answers. It cited about 83 million weekly searches and suggested that better product data could allow Amazon to run those searches using cheaper AI models. Amazon told Reuters the usage figures were outdated and said its own data puts the assistant's global accuracy at about 97 percent. CEO Andy Jassy said more than 400 million people had used it over the past year.

According to people familiar with the project, Amazon executives are concerned that rival AI agents could direct shoppers to retailers with better product information. A more comprehensive catalog could give those agents a reason to search Amazon first, helping protect the company's advertising business. The draft lists October 1, 2027, as a target date, though Reuters could not establish whether or when the project would launch. Amazon described the document as a brainstorming exercise that may not reflect current plans.