Amazon Plans a 'World Product Graph' — a Database of Every Product on the Internet
Amazon is planning a database covering every product on the internet, including those it doesn't sell itself. The World Product Graph project is meant to keep shoppers from leaving for reviews on Reddit and Google, and to give AI agents a ready-made data source. The data would also help Rufus, Amazon's shopping assistant, make fewer mistakes and run on cheaper models. But this is still a draft: Amazon says the document does not reflect current plans.
AI-processed from The Verge; edited by Hamidun News
When you choose headphones or a coffee machine, where do you look for an honest opinion? Most likely not on the store's product page, but on Google or Reddit. Amazon wants to change that. According to an internal company document, it is preparing the World Product Graph, a unified database of information on every product sold anywhere on the internet, including products Amazon does not sell itself.
For developers and product managers, the story matters less as a rumor about a new feature and more as a signal of what large platforms are now competing for: control over which data becomes the primary source when purchases are made by AI agents rather than people.
What Amazon Has in Mind
The idea is simple to illustrate. If you are looking for a specific vacuum cleaner model today, its specs, prices, and reviews are scattered across dozens of sites. You open several tabs, compare them, and then check Reddit to see whether people complain about the model. The World Product Graph is meant to gather this information on Amazon's side: descriptions, parameters, and product data from different stores in one place.
The goal is to make shoppers less likely to leave Amazon to research a product. Even if someone is looking for an item sold by another retailer, they would research it on Amazon's pages. Links to products from external stores already appear there, thanks to a feature the company launched last year.
Why It Matters for AI Agents and the Assistant
The second part of the logic concerns bots. If Amazon ends up with the largest dataset on products, AI agents tasked with choosing and buying an item could rely more on its index than on anyone else's. Then even people who shop through a bot would end up on Amazon's pages.
There is also an internal reason. Amazon has its own AI shopping assistant, Alexa for Shopping, formerly known as Rufus. The document reportedly states plainly that "many Rufus answers are factually wrong or incomplete," and that processing users' search queries with AI is expensive. The reasoning is that the more high-quality data is available, the more accurate the answers and the cheaper the models needed to run them. This follows a familiar principle for RAG systems: a complete and clean knowledge base often delivers more than switching to a more expensive model.
What Remains Uncertain
It is important not to treat the draft as a finished product. First, it is unknown whether the project will launch at all. The document gives a release date of October 1, 2027, but, as the outlet notes, it is unclear whether and when the plan will be implemented. Second, the document itself acknowledges that the project depends on figuring out how brands structure catalogs on their own sites. This may mean Amazon would have to crawl other retailers' websites, which carries technical, legal, and business risk: stores are not obliged to agree to such data collection.
Amazon itself said in a statement that draft and outdated documents often give an incomplete picture, and that the existence of a brainstorming document does not necessarily reflect current plans.
What This Means for Practitioners
If you are building a product for e-commerce or a shopping agent, it is worth considering a few things in advance:
- Structured data on your site is now a question not only of SEO but also of whether you appear in other companies' indexes used by agents. Clear fields, understandable specifications, and up-to-date prices raise the chance that an agent reads your product correctly.
- Do not assume a single aggregator will become the sole source. If Amazon genuinely builds such an index, agents' dependence on it will become a point of competition with Google and other players.
- Data quality affects the cost of AI features. Amazon's admission that Rufus often gets things wrong and is expensive to run is a useful benchmark: errors in answers are often fixed not by a new model but by improving the database the model draws facts from.
For now this is more a bid for the future than an announcement. But the direction is clear: the next round of competition for shoppers is over whose catalog becomes the reference for people and for their AI assistants.
- Nov 21, 2026AG-UI 1.0 gives AI agents a common language for talking to apps: 31 event types
- Nov 21, 2026OpenAI launches ChatGPT Wallet: agents buy without ever seeing your card number
- Nov 21, 2026US Advertisers Spent $70.7 Million on ChatGPT Ads in October, Up 127% From April
- Nov 21, 2026ONNX Runtime vanishes from GitHub: Microsoft hides repositories after package hack
Key Questions Answered
What Amazon Has in Mind?
The idea is simple to illustrate. If you are looking for a specific vacuum cleaner model today, its specs, prices, and reviews are scattered across dozens of sites. You open several tabs, compare them, and then check Reddit to see whether people complain about the model. The World Product Graph is meant to gather this information on Amazon's side: descriptions, parameters, and product data from different stores in one place. The goal is to make shoppers less likely to leave Amazon to research a product.
Why It Matters for AI Agents and the Assistant?
The second part of the logic concerns bots. If Amazon ends up with the largest dataset on products, AI agents tasked with choosing and buying an item could rely more on its index than on anyone else's. Then even people who shop through a bot would end up on Amazon's pages. There is also an internal reason. Amazon has its own AI shopping assistant, Alexa for Shopping, formerly known as Rufus.
What Remains Uncertain?
It is important not to treat the draft as a finished product. First, it is unknown whether the project will launch at all. The document gives a release date of October 1, 2027, but, as the outlet notes, it is unclear whether and when the plan will be implemented. Second, the document itself acknowledges that the project depends on figuring out how brands structure catalogs on their own sites.
Ship AI Faster with Founders, Engineers & Weekly Architecture Reviews
Inside Vibe Club (hamidun.com/club) — weekly live masterminds with Zhemal Khamidun, Tech Radar, full LMS access, and 370+ production Claude Code skills.
The AI world, distilled — once a week
Seven stories that actually mattered, hand-picked. No noise, no reposts, no press releases.
Done! Check your inbox for a confirmation.