SAN FRANCISCO, California: Amazon is drawing up plans to build a database of virtually every product for sale on Earth, whether it stocks the item or not, in a bid to keep shoppers from leaving its site and to blunt the threat from AI shopping agents, according to an internal document reviewed by Reuters and people familiar with the matter.

The initiative, called World Product Graph, could help Amazon cut in to Google's dominance in online search while also fending off rival Walmart's growing e-commerce business.

"Amazon cannot be the go-destination for every shopping mission if customers must leave to research product relationships on Google, brand websites, and review sites," according to the document.

Reuters could not determine if or when this plan may be implemented, though the document, written like a press release, has an aspirational future release date of October 1, 2027. The people said World Product Graph could be modified or aborted due to financial or other concerns.

"Draft, outdated documents often paint an incomplete picture and the existence of a brainstorming document such as this one doesn't necessarily reflect our current plans," said Amazon, in a statement. "We're constantly working on new ideas and technologies to make Amazon's product catalog comprehensive, accurate, and helpful for customers."

World Product Graph "depends on discovering how brands organize their products on their own websites," the document says. That will entail some brands and retailers contributing product data continuously to Amazon, but also Amazon independently crawling other retailers' websites. Reuters could not determine how Amazon planned to crawl the sites or whether it would need to seek permission in all cases.

Amazon has already signed up Coca-Cola and Bayer, among others, to contribute product data, the document says. And the company held a "CEO-level P&G/Amazon summit" for a trial with Procter & Gamble. The companies didn't respond to requests for comment.

The project would also tackle what the document said had been problems with Amazon's AI-powered search engine known as Rufus, which the company began calling "Alexa for Shopping" in May. "Many Rufus answers are factually wrong or incomplete, but customers lack expertise to detect these errors," the document states. Users conduct about 83 million searches weekly using Rufus, and running them is "expensive," according to the document.

In its statement to Reuters, Amazon, referring to the document, said "the usage figures are significantly outdated" and that its data shows that Alexa for Shopping is roughly 97% accurate globally.

World Product Graph would also help stave off the growing threat of alternative AI shopping agents, the people said. CEO Andy Jassy sought during the company's third-quarter earnings call to downplay agents' impact, saying "few customers are doing much shopping this way yet" and that Amazon's advertising business has been minimally impacted.

Yet Amazon executives have grown wary of AI shopping agents that may send customer traffic to other sites that have greater or more complete data, according to people familiar with the matter. Meta's Muse agent has garnered more than 10 million users in the US and Canada since launching in September. Amazon in the same month blocked Muse from its retail site and last year sued AI firm Perplexity to prevent similar access, a case that is ongoing.

By creating the world's most extensive storehouse of product data, Amazon could ensure AI agents working on behalf of shoppers will be compelled to seek products out on Amazon.com and not other retailers' sites, the people said. That would also help protect Amazon's growing advertising business.