Amazon is exploring an ambitious initiative known as the “World Product Graph,” which aims to establish a comprehensive database that encompasses a wide variety of global products. This endeavor seeks to incorporate products even if they are not sold through Amazon’s platform. According to internal documents and insights from knowledgeable sources, the primary objective is to collect detailed information on nearly every product available for sale. This includes aspects such as product specifications, model evolution, colors, and the relationships between different items.
As a company that began with e-commerce, Amazon is taking strides to gather and organize information on non-owned products. The rationale is clear: Amazon wants to create a shopping environment where consumers can access sufficient data for research and comparison without needing to navigate away to platforms like Google or Reddit to make informed purchase decisions. In essence, Amazon envisions itself not only as a platform for placing orders but also as a primary repository for product research and comparison prior to making a purchase.
Many shoppers assume that online retail is synonymous with visiting Amazon. However, consumer behavior reveals a different story, as various search platforms are utilized at different stages of the shopping journey.
A 2025 research report from Morgan Stanley indicates that when consumers have already decided on a product, 41% initiate their search on Amazon, while Google and Walmart account for 23% and 18%, respectively. In contrast, during the initial phases of research and comparison—before finalizing product choices—57% of consumers begin with Google, with only 16% opting for Amazon.
This discrepancy is precisely what Amazon aims to change. The internal documents indicate the company desires to retain consumers on its platform by providing comprehensive product data that facilitates product research, comparison, and purchasing.
Another pressing context for this initiative is the rise of AI shopping agents. In the near future, consumers may no longer browse websites manually; instead, they will employ AI to search for, compare products, and even finalize their purchases.
As mentioned in internal communications, Amazon has begun collaborating with major companies, including Coca-Cola and Bayer, to gather product data. The company has also conducted “CEO-level summits” with Procter & Gamble to foster pilot collaboration. Furthermore, Amazon plans to scrape data from other retail websites to enhance its product information database. However, the specifics of this data scraping method and whether Amazon would require permissions from individual websites remain uncertain.
External AI shopping tools have begun to pose competition for Amazon. Since its launch in September 2026, Meta’s Muse shopping agent has attracted over 10 million users in the U.S. and Canada. In response, Amazon blocked Muse on its retail website the same month and has ongoing litigation against AI firm Perplexity to prevent similar access.
Last month, Elon Musk addressed the news of Amazon blocking his xAI-run Grok agent on the X platform, suggesting that perhaps it was time to found a rival to Amazon.
Amazon is concerned that if external AI systems gain access to more comprehensive product data, they could divert consumer traffic and advertising revenues away to other retailers. According to insiders, Amazon is keen on establishing a global database of products to ensure that AI agents searching for products in the future depend on Amazon’s platform, thereby safeguarding its e-commerce and advertising business.
Additionally, this initiative aligns with Amazon’s own AI shopping tools. Internal documents revealed that many responses from Amazon’s AI shopping assistant, renamed Alexa for Shopping earlier this year, contained errors or incomplete information. Consumers, lacking expertise, might not catch these inaccuracies.
The documents highlight that Rufus, as it was previously known, receives around 83 million searches weekly, resulting in high operational costs. Developing a more extensive product database could enhance the efficiency of lower-cost AI models. However, Amazon stated that the usage data referenced in the documents is significantly outdated, asserting that current figures indicate a 97% accuracy rate for Alexa for Shopping globally.
One passage within the internal proposal captures a hypothetical customer’s sentiment: “I used to spend an hour on Reddit researching these specifications, but now it feels like this store truly knows what it’s selling.”
This document belongs to a category of internal proposals at Amazon that evaluate new product concepts in the format of PR/FAQ drafts, which include press release styles and common queries. Although the proposed launch date is set for October 1, 2027, this does not imply that the project has received final approval.
In response to inquiries, Amazon emphasized that brainstorming documents often lack completeness and do not necessarily represent the company’s current plans. Furthermore, the timeline for a formal rollout of this initiative remains unclear.
An insider revealed that Amazon is adjusting its personnel allocation to push the project forward, but noted that the “World Product Graph” could still be subject to modifications or even cancellation due to financial considerations or other reasons.