Amazon is planning an initiative called World Product Graph to index global product data, aiming to prevent consumers and third-party AI shopping agents from leaving its site. Internal documents reveal Amazon's Rufus AI assistant faces accuracy issues and high operational costs across 83 million weekly searches. While Amazon pushbacks call the document outdated and report 97% accuracy, the strategy seeks to counter Google search dominance and blocking threats from AI tools like Meta's Muse.
World Product Graph initiative details
- ▪An internal Amazon document lists an aspirational release date of October 1, 2027 for the World Product Graph initiative
- ▪An internal Amazon memo details an initiative called World Product Graph to gather data on virtually every item for sale globally, including products Amazon does not stock
- ▪The World Product Graph initiative is organized under Robert Tekiela, a vice president in Amazon's Selection & Catalog Systems unit
- ▪Companies including Coca-Cola and Bayer signed up to contribute product data to Amazon, while Amazon held a summit with Procter & Gamble for a product data trial
Rufus AI search agent performance and costs
- ▪Expanding Amazon's product dataset through the World Product Graph is intended to allow Rufus to utilize cheaper, more efficient artificial intelligence models
- ▪An internal Amazon document states that many answers provided by Amazon's Rufus AI search agent are factually wrong or incomplete
- ▪An internal document indicates users conduct about 83 million weekly searches using Amazon's Rufus AI agent, making operation expensive
Amazon product data gaps and competitive risks
- ▪Amazon executive Robert Tekiela stated in an internal document that the site lacks an understanding of how products relate to each other and which item suits a customer
- ▪An internal document asserts Amazon cannot remain the primary destination for shopping missions if customers must leave the site to research product details
- ▪Amazon executives expressed concern that external AI shopping agents could redirect customer traffic to competing retail sites possessing more complete product data
External AI shopping agents and Amazon blockades
- ▪Amazon blocked Meta's Muse agent from its retail site in September 2026 and filed a lawsuit against AI firm Perplexity to prevent site access
- ▪Meta's Muse AI shopping agent accumulated more than 10 million users in the United States and Canada following its September 2026 launch
Alexa for Shopping performance and usage
- ▪Amazon stated that internal usage figures cited by Reuters are outdated and claimed its Alexa for Shopping tool is roughly 97% accurate globally
- ▪Amazon CEO Andy Jassy stated that the company's Alexa for Shopping tool was used by more than 400 million people over the past year
Story comments
Loading comments…