Amazon AI Shopping Plan Would Crawl Rivals to Beat AI Agents
An internal Amazon memo targets a 2027 database of nearly every product sold online, after Rufus hit 83 million weekly searches. Here's why it matters.
Hakkı ALKANPublished Updated 4 min read
Amazon has drafted a plan to collect data on nearly every product sold anywhere, including items it does not stock, so shoppers and AI agents never need to leave its store, Reuters reported on Nov. 13, 2026. The project, called "World Product Graph," would combine data supplied by brands with information gathered by crawling rival retailers' websites, and it carries a target date of Oct. 1, 2027.
Amazon told Reuters the document is an outdated brainstorming draft that "doesn't necessarily reflect our current plans," so a launch is not confirmed. Still, the memo shows how worried the company is about AI shopping agents, software that researches products and places orders on a user's behalf.
What Is Amazon's World Product Graph?
A product graph is a database that maps how products relate to each other: which model replaced which, which colorway is current, which accessory fits, and which rival is a close alternative. Amazon wants that map for virtually every item on sale, whether or not it sells the item itself.
According to the internal document, the plan "depends on discovering how brands organize their products on their own websites." Some brands and retailers would feed data to Amazon continuously, while Amazon would also crawl, or automatically scan, other retailers' sites. Reuters could not determine how that crawling would work or whether Amazon would ask permission in every case.
The memo says Coca-Cola and Bayer have already signed up to contribute data, and that Amazon held a "CEO-level P&G/Amazon summit" for a trial with Procter & Gamble. The project sits in Amazon's Selection & Catalog Systems unit under vice president Robert Tekiela.
Why Does Amazon Want Shoppers to Stop Googling?
The document is blunt about the problem: "Amazon cannot be the go-destination for every shopping mission if customers must leave to research product relationships on Google, brand websites, and review sites." The numbers explain the urgency.
| Where online shoppers start (Morgan Stanley, 2025) | Amazon | Walmart | |
|---|---|---|---|
| General product research | 57% | 16% | n/a |
| Already know the exact product | 23% | 41% | 18% |
In other words, Amazon wins when you know what to buy but loses the earlier research phase, where reviews, Reddit threads and brand pages shape the decision. That research phase is exactly where AI assistants now sit.
How Do AI Shopping Agents Threaten Amazon?
Agents can compare stores and send buyers wherever the data is best. Meta's Muse agent has passed 10 million users in the US and Canada since launching in September, and Amazon blocked it from its retail site that same month. Meta is also pushing cheaper agent tasks, as we covered with Muse Ember's roughly $0.14 per task pricing.
Amazon sued Perplexity last year to stop its shopping agent from accessing Amazon, a case that is still ongoing. OpenAI is moving in the same direction, with ChatGPT Wallet letting agents shop for you without your card.
The people Reuters spoke to said the logic is simple: if Amazon holds the most complete product data on the web, agents working for shoppers will be pushed back to Amazon.com. That would also protect Amazon's growing advertising business.
Is Alexa for Shopping Accurate Enough?
The memo also targets the weakest link in Amazon AI shopping today: its own assistant. Rufus, renamed "Alexa for Shopping" in May, handled about 83 million searches a week according to the document, which called those queries "expensive" and said "many Rufus answers are factually wrong or incomplete, but customers lack expertise to detect these errors."
Amazon pushed back, saying the usage figures are significantly outdated and that Alexa for Shopping is roughly 97% accurate globally. CEO Andy Jassy has said more than 400 million people used it in the past year. Amazon has kept adding ways to buy through Alexa, such as ordering by text message in the US.
What Does Amazon AI Shopping Mean for You?
- Shoppers could get model comparisons and "which one is right for me" answers without leaving Amazon.
- Those answers would come from the store that profits from the sale, so cross-checking prices elsewhere still matters.
- Brands may feel pressure to hand structured product data to Amazon directly.
- Smaller retailers could see their pages crawled while traffic flows back to Amazon.
- Third-party agents from Meta, Perplexity, OpenAI or xAI may find Amazon even harder to work around.
The plan builds on existing "Buy for Me" and "Shop Direct" features, which already list products from other sites and let Amazon act as the middleman. You can read the full details in the Reuters report, with added analysis from The Verge.
Our Take
This is a defensive move dressed up as a customer feature. Amazon already wins the "buy" moment; World Product Graph is an attempt to own the "decide" moment before Google, Reddit or an AI agent does. For shoppers, a smarter catalog is genuinely useful, but a store that knows every rival's product data is not a neutral referee. If the 2027 target holds, the next phase of Amazon AI shopping will be a fight over who controls product data, and whether retailers get a say before Amazon's crawlers arrive.