Reuters reported on November 13, 2026 that Amazon has an internal plan, called World Product Graph, to gather data on virtually every item sold anywhere, whether or not Amazon stocks it. The aim, per the report, is to keep shoppers and AI shopping agents on Amazon. Amazon says the document is an outdated brainstorm. No launch is confirmed.
This post lays out what Reuters says it saw, what Amazon answered, and what the plan would mean if it ever ships. It builds on our earlier coverage of Amazon blocking Meta's Muse shopping agent.
TL;DR: reported vs confirmed
| Question | Status |
|---|---|
| Does an internal World Product Graph document exist? | Reported by Reuters, which reviewed it; Amazon confirms a "brainstorming document" exists |
| Is it a product? | No. Reuters could not determine whether or when it will launch |
| Target date in the document | October 1, 2027, described as aspirational |
| Who leads it? | VP Robert Tekiela, Selection & Catalog Systems, per Reuters; he did not respond |
| Brands named as signed up | Coca-Cola and Bayer, per the document; none responded to Reuters |
| Will Amazon crawl other retailers? | The document says so; Reuters could not determine how or whether permission is needed |
| Rufus / Alexa for Shopping accuracy | Document says many answers are wrong or incomplete; Amazon says roughly 97% accurate |
What does the Reuters report say?
The exclusive, by Reuters reporter Greg Bensinger, is republished on Investing.com. Reuters says it reviewed an internal document and spoke with people familiar with the effort. Key points:
- The goal is to gather data on virtually every product sold on Earth. The document says Amazon "cannot be the go-destination for every shopping mission" if shoppers must leave to research products elsewhere.
- The effort "depends on discovering how brands organize their products on their own websites."
- Some brands would contribute data continuously. Amazon would also crawl other retailers' sites.
- The document names Coca-Cola and Bayer as signed up, and mentions a "CEO-level P&G/Amazon summit" for a trial with Procter & Gamble.
- Amazon is reassigning staff to get the project started.
- The memo is a PR/FAQ of about six pages, a format Amazon uses to propose products. Such documents do not always become products.
Reuters adds that the plan could cut into Google's search lead and counter Walmart's e-commerce growth, and cites a 2025 Morgan Stanley report: 16% of online shoppers begin product research on Amazon versus 57% on Google. When shoppers know what they want, 41% start on Amazon.
Why is Amazon worried about AI shopping agents?
Shopping cart stopped at a barrier gate, illustrating Amazon World Product Graph and blocked AI shopping agents
An AI shopping agent researches and buys on a user's behalf. If it finds better product data on another site, it can send the purchase there and bypass Amazon's ads and search. Reuters says executives are wary of agents that steer traffic to sites with more complete data, and that a stronger catalog could point agents back to Amazon.com and protect its growing advertising business.
At the same time, Amazon CEO Andy Jassy said on the Q3 earnings call that "few customers are doing much shopping this way yet," and that advertising has been minimally affected, per Reuters.
Amazon has already acted against outside agents. Reuters says it blocked Meta's Muse agent in September, a move we covered in our Muse blocking post, and that its lawsuit against Perplexity over similar access is ongoing. Reuters also cites JPMorgan saying Muse "has the potential to become the most widely used consumer AI application since ChatGPT," and notes Elon Musk wrote last month, "Maybe we should start a competitor to Amazon," after a post said xAI's Grok agent was blocked. Our earlier piece on Paul Graham calling the agent ban a startup opportunity covers that reaction.
What does the memo say about Rufus?
Amazon's shopping assistant Rufus was renamed "Alexa for Shopping" in May, per Reuters. The document reportedly says many of its answers are wrong or incomplete, and that customers often cannot spot the errors. It also says users run about 83 million Rufus searches a week, which the document calls "expensive," and that richer product data would let Amazon use cheaper models.
Amazon pushed back on this part directly. It said the usage figures are "significantly outdated" and that Alexa for Shopping is roughly 97% accurate globally. Reuters also reports Jassy said Alexa for Shopping has been used by over 400 million people in the past year. The accuracy dispute is unresolved: Reuters is describing an internal document, Amazon is giving a company-wide figure, and neither has published its method.
Claimed vs verified: what we can and cannot say
Confirmed by Amazon: a brainstorming document exists, and Amazon is "constantly working on new ideas and technologies" to make its catalog comprehensive, accurate and helpful. Its words: "Draft, outdated documents often paint an incomplete picture."
Reported, not independently confirmed: the World Product Graph name, the October 1, 2027 target date, the Coca-Cola, Bayer and P&G involvement, the Rufus search count, and the crawling plan. We read the story through a republished copy of the Reuters text, not the document itself.
Unknown: whether crawling would respect robots.txt or site terms, how brands would opt in, how much it would cost, and whether it is still alive.
How does this relate to Buy for Me and Shop Direct?
Employees quoted by Reuters describe World Product Graph as a step toward a one-stop shop for buying anything. It would expand "Buy for Me" and "Shop Direct," features that list products sold on other sites and let customers buy them through Amazon. Reporting from earlier in 2026, which we did not verify independently, says some small brands found Amazon listings of their products that they had not authorized, with AI-written descriptions that did not match. A bigger catalog built by crawling would raise the same question at larger scale.
In the document, Tekiela is quoted as saying customers can already find products on Amazon, but what is missing is understanding how products relate and which one fits them. That is a data-modeling problem: attributes, compatibility, and comparisons, not just listings.
What this means for brands, sellers and agent builders
Small cart with a green price tag, showing product data that shopping agents read
- Brands. The memo suggests Amazon wants structured feeds from brands. If you sell on your own site, check how your product pages expose attributes, because crawlers and agents read them. Clean structured data helps every agent, not only Amazon's.
- Sellers on Amazon. Watch for listings built from your off-Amazon pages. Review your Amazon catalog regularly for descriptions or images that do not match your products.
- Agent builders. The conflict is about access. Platforms that hold catalogs and checkout can block outside agents, so plan for official APIs, partnerships or user-authorized sessions. Our guide to AI shopping answers and risky sites shows how accuracy varies across agents, and Google AI Mode shopping prices shows another angle on agent-led commerce.
- Payments. Agent checkout is also being built from the payments side; see Stripe Link and Grok shopping.
How should readers weigh a leaked planning memo?
Amazon's PR/FAQ format is worth understanding. Teams write a mock press release and a list of frequently asked questions for a product that does not exist yet, then argue over it. The point is to test an idea on paper before building it. A memo in this format shows what a team wants to pursue and how it frames the case. It does not show a funded roadmap.
Three signals in the Reuters account suggest the idea has some weight: a named leader inside the catalog organization, staff being reassigned to start the work, and named brand partners plus a trial with a large consumer-goods company. Three signals point the other way: an aspirational date nearly a year away, Amazon's statement that the document is outdated, and Reuters' note that the plan could be changed or dropped over financial or other concerns.
The accuracy dispute is also a reminder to separate two kinds of numbers. A figure inside a memo, such as weekly search volume, was written to make an internal argument. A figure Amazon states publicly, such as 97 percent accuracy, was chosen for public messaging. Neither comes with a published test method, so readers should treat both as claims. If you are deciding whether to rely on an AI shopping assistant, test it yourself on products you know well, and check that the price, size and compatibility it states match the seller's page before buying.
What to watch next
- A formal Amazon announcement. Any official launch of a cross-web catalog would settle the status. Today there is none.
- Crawling terms. A public statement on robots.txt, permissions or brand opt-in would tell brands whether this is a partnership or scraping.
- Court and policy moves. The Perplexity case and any new access disputes show how far platforms can restrict agents.
- Brand deals. Confirmation from Coca-Cola, Bayer or P&G would show the plan has real partners.
Until then, read this as a window into Amazon's thinking, not a product. The report is a strong signal that the biggest retailer sees catalog data as its main defense against shopping agents, and the company has not said it will build it.
Related reading
- Amazon blocks Meta Muse shopping agent
- Paul Graham: Amazon banning agents is a startup opportunity
- AI shopping answers and risky sites, claimed vs verified
- Google AI Mode shopping prices
- Stripe Link and Grok shopping
- Amazon seller assistant Claude plugin
- Sources: Reuters exclusive via Investing.com, eWEEK on retailer backlash to Amazon scraping
Details reflect Reuters reporting and Amazon's response as of November 13, 2026 and may change.