Amazon is drawing up plans for a “World Product Graph,” a giant database of every product sold anywhere on the internet, including products Amazon doesn’t sell itself. The Verge AI reports, citing Reuters, that the company wants to surface all of that information on its own site. An internal document puts the target release date at October 1st, 2027. Amazon calls the document a brainstorm, so it’s not a confirmed plan yet.
The Short Version
- What: A product index covering the whole web, not just Amazon’s own catalog.
- Why: Keep shoppers from leaving Amazon to research on Reddit or Google, and make Amazon the default data source for AI shopping agents.
- The AI angle: It would feed Amazon’s shopping assistant, Alexa for Shopping (formerly Rufus). An internal document admits that assistant has accuracy problems.
- Status: Unconfirmed. Amazon says draft documents “often paint an incomplete picture.”
Amazon Is Playing Defense
Amazon’s problem is that product research mostly happens somewhere else. People check Reddit threads, Google results, and brand websites before they buy. Every one of those detours is a chance for the sale to go to another store.
The World Product Graph tries to close those exits. If Amazon has data on every product, shoppers have less reason to leave, even when the thing they want is sold elsewhere. Amazon already started down this road last year with its “shop direct” feature, which links out to other retailers’ products from inside Amazon.
The Real Target Is AI Agents
This matters most for agentic commerce. AI agents are starting to shop for people, and they pull from whatever product data is biggest and most reliable.
If Amazon owns the largest product index, agents will probably lean on it. So even people who never open Amazon themselves could end up on its pages through a bot. Search engines built moats by indexing the web. Amazon looks like it wants to do the same for shopping, just as agents become a real way to buy things.
Amazon Admits Its Assistant Gets Things Wrong
The most revealing detail is how frank the internal document is. According to Reuters, it says “Many Rufus answers are factually wrong or incomplete.” It also says running customer searches through the AI costs a lot.
Amazon thinks more data fixes both problems:
- Accuracy: Better structured product information means the assistant guesses less and hallucinates less.
- Cost: With good data to pull from, Amazon could answer questions with smaller, cheaper models instead of expensive frontier ones.
That’s a common lesson across the industry. Retrieval quality often matters more than model size. A mid-sized model with clean, structured data can beat a giant model that’s working from memory, and it costs far less to run at Amazon’s scale.
The Hard Part: Crawling Competitors
The document says the project “depends on discovering how brands organize their products on their own websites.” Put simply, Amazon may need to crawl other retailers’ sites to build the index.
That raises some obvious questions:
- Retailer pushback: Competitors may not want to supply the data that pulls their customers into Amazon.
- Blocking: Sites can and do block crawlers, and many already restrict AI scrapers.
- Regulators: A dominant marketplace indexing its rivals’ catalogs will likely get antitrust attention.
What to Watch
For retailers and brands, the takeaway is simple: product data is now a strategic asset. If AI agents decide where purchases happen, clean and well-structured product information on your own site decides whether you show up.
For AI builders, this shows where the shopping-agent race is heading. Everyone is fighting over data, not just models. Expect Google, OpenAI, and Shopify to keep building their own product indexes in response.
Reuters says it’s unclear “if or when” Amazon will build this. Even so, the direction is clear. Amazon wants to be where both shoppers and their AI agents start. You can find more detail in the original report from The Verge AI.