Amazon has plans to build a vault of data on every product sold anywhere on the internet, according to Reuters. That includes products Amazon doesn’t sell. The project is called the “World Product Graph”. The information would be made available through Amazon’s own site, and the aim is to make it less likely that shoppers leave Amazon to do their research. The Verge reported on it on 13 November 2026.
I could only read the feed summary of the story, not the full article. So I am keeping to what that summary says and not guessing at the details.
The excitement is about Amazon’s reach
The easy reading is that Amazon is now the reference book for all of retail. A shopper asks a question, and Amazon answers it, even about a product from a shop Amazon has never heard of.
Is there a real need? For Amazon, yes. The stated goal is to keep shoppers from leaving to research elsewhere. For shoppers, a single place that knows every product is useful, if the information is right.
What changed is where the product facts sit. Until now, a shopper compared products across many sites. Under this plan, a copy of your facts sits inside someone else’s system, and it may be summarised for a shopper you never see.
What didn’t change is that the index has to get its facts from somewhere. It reads what is published. I don’t know how Amazon will collect the data, and the summary doesn’t say. Whatever the method, the source material is the public web, and your own pages are the part of it you control.
Your facts will be read somewhere you don’t control
I hold a position on agentic commerce. It is too early to call, and I would put it at 50/50. Some platforms will open up and welcome agents. Others will restrict access to keep their grip on shoppers. The big platforms can afford to wait and see, or to shape how it turns out.
Smaller businesses can’t. For them the shift is already decided, and they need to get data-centric. Your website is one surface of many. Wherever your data shows up, it needs more precision and more detail, and it needs to say accurately what you are.
The Amazon project is a clear example. If a product graph holds a record of what you sell, that record is built from whatever was easiest to find and read. A vague record gets summarised vaguely. A competitor’s cleaner record gets the recommendation.
What a precise record looks like
Take a small coffee roaster selling beans online. Its product page says “rich, smooth, premium beans, roasted with care”. Those are adjectives. Any index that reads that page learns nothing it can quote, and nothing that separates this roaster from fifty others.
Now take the same page with a roast date policy, the origin and altitude, the roast level, the bag sizes in grams, and the delivery time in days. Those are attributes that are measurably different from a competitor’s. An index can quote them. A shopper’s agent can compare them.
Numbers survive and adjectives don’t. Engines quote specific facts, and marketing language gets dropped. That holds for an AI search engine, and I expect it holds for a product graph too.
The same applies to a clinic listing treatments, a service business listing what is included in each package, or a marketing team looking after a catalogue of hundreds of items. The more attributes you publish that are specific and different from your competitors’, the better the odds that any index gets your brand right.
Then there is the long tail. Everyone asks their agent the same question in slightly different words. To catch that, your information has to cover a wider spread of questions, with varied and detailed attributes. We do this at AeroChat, which publishes many pages targeting individual integrations with specific software. Each page catches one long-tail question.
What I would do this month
I wouldn’t try to work out how the World Product Graph collects its data. Nobody outside Amazon knows yet, and the answer may change. The strategy holds either way, so I wouldn’t spend effort on tactics.
Here is what I would do.
- Pick your ten best-selling products or services. Read each page as a stranger would. Mark every adjective that has no number or fact behind it.
- Replace each marked line with something measurable: sizes, weights, times, materials, what is included, what is excluded, what it costs.
- Add the attributes that make you different from your closest competitor. If you can’t name any, that is the more urgent problem.
- Write more pages for the specific questions buyers ask, one question per page, and keep the facts consistent with your product pages.
- Check what AI search engines and shopping assistants say about you, and note the date. Answers change week to week, so repeat the check.
Your own pages are the record you write. Any index that collects your facts will start from them.
Source: The Verge AI: Amazon has a plan to index the world’s online shopping data