Amazon's 'World Product Graph' Would Index Every Product Online
Amazon's internal 'World Product Graph' plan would index every product sold online, with a document listing October 1, 2027 as the release date, Reuters reports.

- An internal Amazon document lists October 1, 2027 as the World Product Graph release date, per Reuters.
- The document states 'Many Rufus answers are factually wrong or incomplete' and that AI search is expensive to run.
- Amazon launched its 'shop direct' feature linking to outside retailers' products last year.
- Amazon said the document is a brainstorming draft and 'doesn't necessarily reflect our current plans.'
- The project depends on discovering how brands organize products on their own websites, the document says.
Amazon is planning to build a product index covering everything sold anywhere on the internet — including items it doesn't sell itself — with an internal document pegging a release date of October 1, 2027, according to a Reuters report.
The project is called the World Product Graph. Its goal is simple: keep shoppers from leaving Amazon to research products on Reddit or Google. If Amazon holds the richest product dataset on the web, the thinking goes, shoppers will end up on Amazon's pages even when the product they want is sold elsewhere. The same logic applies to AI shopping agents — bots that help people buy things will lean on the largest index available, and Amazon wants that index to be its own.
The move reads as a direct play for the kind of product research traffic that currently flows through Google search and Reddit discussions. It also extends a feature Amazon already runs: "shop direct", launched last year, which surfaces links to outside retailers' products on Amazon's own site.
What does the plan fix?
Amazon's own AI shopping assistant, Alexa for Shopping — previously called Rufus — is a central part of the story. The World Product Graph would feed the assistant more data to make it more accurate and cheaper to operate.
The internal document is blunt about the current state of the tool. It says: "Many Rufus answers are factually wrong or incomplete." It also notes that running users' Amazon searches through the AI is expensive. More data, the document reportedly argues, would let the system use cheaper models while improving accuracy — a swap that could cut the cost of every AI-assisted search.
That detail matters beyond Amazon. It's a concrete signal that large AI shopping systems live or die on the breadth of the data underneath them, and that compute cost, not just answer quality, drives product decisions at this scale.
What's actually confirmed?
Here it's worth separating what's solid from what's speculative:
- Confirmed by Amazon: the company acknowledged the document exists. In a statement to Reuters, Amazon said: "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."
- Reported, not confirmed: the project scope, the October 1, 2027 date, the Rufus criticisms, and the cost rationale all come from an internal document Reuters reviewed. Amazon has not committed to shipping any of it.
- Unknown: Reuters says it's unclear "if or when this plan may be implemented."
Amazon's statement is a soft denial of nothing in particular. It doesn't dispute the document's contents; it disputes that a brainstorming document equals a roadmap.
What stands in the way?
The internal document names the core obstacle itself. The project, it says, "depends on discovering how brands organize their products on their own websites." That could mean Amazon crawling other retailers' sites to map how they structure and present their catalogs — a technical and potentially legal challenge that echoes how Google's crawler built the foundation of its search index.
Crawling at that scale raises questions the report doesn't answer. Will retailers cooperate, block, or litigate? The document itself doesn't say, and neither does Amazon.
Why should search watchers care?
For anyone tracking where product discovery happens, this is a competitive threat aimed at two fronts at once: Google's product-search dominance and Reddit's growing role as a trusted research layer. A comprehensive product index plus an AI assistant would give Amazon an answer engine for shopping that sits on its own shelf — and AI agents acting on behalf of shoppers would land there too, by default of data volume.
The takeaway in plain words: Amazon wants to become the index of record for everything for sale online, and it has a dated internal document saying so — but no public commitment yet.
Watch for signals over the coming months: expansion of the shop direct feature, Rufus accuracy improvements, any crawler activity from Amazon hitting retailers' sites, and whether the 2027 date survives contact with reality.
via theverge.com (Source)
More from Olivia Hart
Show full bio
Staff writer covering money, funding and market shifts around Search rankings and algorithm updates, generative engine optimization at SERP Journal.
216 articles