Amazon is developing a proposed World Product Graph covering products sold worldwide, including those it does not carry, with an aspirational launch date of 1 October 2027, Reuters reported on 13 November 2026. An internal planning document describes a database intended to improve Amazon’s catalogue and shopping assistant, and to compete with AI agents that can direct shoppers to other retailers.
Key points
- The proposal combines product feeds from brands and retailers with automated collection from rival websites.
- Coca-Cola and Bayer have signed up to provide data, according to the document. Procter & Gamble attended a trial summit.
- Amazon says the planning memo is outdated and puts Alexa for Shopping’s global accuracy at roughly 97%.
World Product Graph reaches beyond Amazon’s catalogue
The proposed database would bring together information supplied by brands and retailers and information collected by crawling competing retailers’ sites. That second method would let Amazon gather product details directly from websites rather than relying solely on businesses to supply feeds. The document’s scope extends to items a shopper could research on Amazon but could not buy there.
The document says Coca-Cola and Bayer have signed up to contribute product data. Procter & Gamble attended a trial summit connected to the effort. The distinction matters to the project’s progress: the document describes the first two companies as data contributors and the third as an attendee at an early meeting.
The plan sets a public release target rather than a fixed delivery date. Reuters reported that the project could be changed or abandoned. Amazon told the news agency that the document was an outdated brainstorming memo and did not necessarily represent its current plans.
Amazon’s search position faces AI agents
Research figures included in reporting on the plan put Amazon’s share of direct product searches at 41%, against 23% for Google and 18% for Walmart, Crypto Briefing reported. The proposed graph would cover a wider range of goods than Amazon’s own inventory, addressing shoppers who otherwise look elsewhere for product information.
The document identifies Google and other destinations for product research as competitors for shoppers’ attention. AI shopping agents add another route: they can browse and compare products on a user’s behalf and potentially complete a purchase. A catalogue that includes goods sold by other retailers would give Amazon’s own assistant information for questions about products outside its store.
Amazon has also acted against an outside agent operating on its marketplace. It sued Perplexity in November 2025 over the Comet browser agent, accusing it of presenting automated visits as ordinary Chrome traffic. A preliminary injunction granted in March 2026 was vacated by a federal appeals court in August, while the underlying case continued.
Alexa for Shopping’s accuracy and running cost
The planning document also sets out a problem within Amazon’s own shopping tools. It says many answers from Rufus, the assistant renamed Alexa for Shopping in 2026, are wrong or incomplete. According to the document, about 83 million searches a week ran through the assistant at high cost. Its case for richer product data includes the prospect of answering those questions with less expensive models.
That assessment is disputed by Amazon. The company told Reuters that the usage figures in the memo were significantly outdated. It said its own data puts Alexa for Shopping’s accuracy at roughly 97% globally.
