Amazon has plans to build a vault of data on every product sold anywhere on the internet, even products Amazon doesn’t sell, and make the information available through its own site, according to Reuters. Labeled internally as the World Product Graph, the project is meant to make it less likely that shoppers leave Amazon to do their product research on Reddit or Google, and make it more likely shoppers end up at Amazon even if they’re searching for a product sold elsewhere. The initiative marks a structural effort to make Amazon the primary destination for queries even when transactions occur on outside platforms.
Amazon seeks to ensure autonomous shopping agents rely on its index of products over others, so even people shopping with bots will end up on its pages. By establishing a comprehensive catalog of global digital inventory, Amazon intends to position its database as the foundational layer for automated commerce transactions.
Initial steps toward indexing external inventory are already visible in Amazon's public marketplace interface. Links to outside retailers’ products have already started popping up on Amazon through the “shop direct” feature it launched last year.
Rufus Assistant Flaws and Costly Queries
Amazon has its own AI shopping assistant, Alexa for Shopping, previously called Rufus. The engineering architecture behind the assistant faces direct challenges regarding operating costs, resource allocation, and inference accuracy. In the current economic landscape of large language models, computing inference at the scale required for global ecommerce queries imposes a severe capital drain on operating margins.
An internal Amazon document on the project says.
"Many Rufus answers are factually wrong or incomplete"
According to the internal document, running users’ Amazon searches with the AI is expensive. The high cost of processing complex natural language queries through advanced models creates a distinct inference cost deficit that strains unit economics. Adding more structured data to the underlying index is reportedly supposed to make the system more accurate and able to route requests to cheaper models, thereby lowering the per-query computation expenses that currently threaten profitability.
How will Amazon acquire the structured data needed to run cheaper models?
Data Collection Challenges and Launch Timelines
Acquiring external merchant product architectures introduces immediate data-gathering obstacles. The technical barriers of crawling external merchant sites stem from the vast fragmentation of website structures, non-standardized taxonomies, and dynamic anti-scraping defenses deployed by competing retailers. An internal document reportedly says the project “depends on discovering how brands organize their products on their own websites,” which requires Amazon to navigate these deep architectural inconsistencies across thousands of disparate platforms.
These technical hurdles directly impact the projected October 2027 timeline, as parsing unstructured or heavily guarded external catalogs demands extensive engineering iterations to normalize the data into a cohesive graph.
Amazon said in a statement to Reuters 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."
Reuters says it’s unclear “if or when this plan may be implemented,” but said the document on it included an October 1st, 2027 release date.
For comparison
As buyers drift away to Reddit or Google to research products, running complex language models over broad queries eats into operational margins. The company aims to index competitor catalogs simply to feed automated shopping agents with cheaper inference directly inside its storefront. Turning an entire ecosystem into the default infrastructure for automated commerce now hinges on parsing guarded third-party catalogs by October 2027.