Every major platform is closing its data loop. Amazon just closed two at once.

On July 8, 2026, Amazon replaced specific product names in order confirmation emails with generic category labels. “1 Electronics item.” “1 Household item.” “1 Drugstore item.”

A developer tracking the transition found the last detailed email arriving at 12:56 PM ET; the first generic one arriving six hours later.

The categories are often wrong. Bloodstones labeled as Drugstore items. A torch lighter, also Drugstore. A garden statue labeled Decor. That inaccuracy is informative; getting the categories right was never the goal. Amazon has issued no public explanation. The silence tells you more than a statement would have.

TL;DR

• Amazon removed product names from order confirmation emails, replacing them with vague category labels. No opt-out.
• Gmail has built purchase tracking on order email data since at least 2019. Its dedicated Purchases tab launched September 2025. Amazon just made it useless for its own orders.
• Simultaneously, Amazon has blocked every major AI shopping agent from crawling its site: ChatGPT, ClaudeBot, Perplexity, Google’s Project Mariner, Meta… while building its own AI shopping assistant.
• One strategy: Amazon is defending its transaction data against both legacy search and the AI agents replacing it.
• For brands: every platform closes its loop eventually. The asset that survives is structured, server-rendered, on-site content on your own domai .

What Google Lost (and When It Built It)

To understand what closed on July 8, you need to understand how much was open before it.

Since at least 2019, Gmail automatically parsed order confirmation emails to build purchase records for its users. A CNBC investigation that year found every Gmail account had an attached purchase history compiled silently from digital receipts, without any explicit disclosure. This ran despite a 2017 commitment from CEO Sundar Pichai to stop using Gmail content for ad targeting.

On September 11, 2025, Google launched a dedicated Purchases tab in Gmail, rolling out globally across Android, iOS, and web. The tab consolidated purchase emails and package tracking into one view. Google called it convenience. It was also the most visible surface of a data collection operation built on other companies’ transactional emails.

Third-party email apps were doing the same thing at commercial scale. A February 2020 VICE/Motherboard investigation found that Edison Mail, Slice (owned by Rakuten), and Cleanfox were scraping inboxes and selling aggregate data to finance, travel, and ecommerce clients. A J.P. Morgan document described Edison’s product as providing “consumer purchase metrics including brand loyalty, wallet share, purchase preferences.” Slice reportedly priced access to individual data categories at over $100,000. Edison marketed itself as “Privacy First.”

Amazon’s order confirmation emails were the highest-value input for all of these products. Amazon handles a large share of US ecommerce volume. Parsing those emails gave Google, Edison, and everyone else a near-real-time view of US consumer purchasing at the brand and category level. That view is now blank.

The Second Front: AI Shopping Agents

The email change is significant on its own. Amazon is running the same playbook on a second front simultaneously, and that one matters more for where commerce is heading.

Since mid-2025, Amazon has systematically blocked AI shopping agents from crawling its site. By August 21, 2025, Amazon’s robots.txt blocked Anthropic’s ClaudeBot, Perplexity’s crawler, Google’s Project Mariner, Meta, Huawei, and Mistral. By November 2025, it had added OpenAI’s ChatGPT-User and OAI- SearchBot on top of the already-blocked GPTBot. Amazon also pulled its shopping inventory out of Google Shopping ads.

The commercial logic is direct. AI shopping referrals to US ecommerce sites grew 752% year-over-year during the 2025 holiday season. By August 2025, ChatGPT alone accounted for nearly 21% of Walmart’s incoming referral traffic, over 20% of Etsy’s, and roughly 15% of Target’s. Generative AI became the second-most popular source for shopping recommendations in 2025, behind physical stores and ahead of social media and brand websites.

If an AI agent can browse Amazon’s catalog and check prices without sending a customer to Amazon.com, Amazon’s retail traffic, and the advertising revenue attached to it, erodes. Every shopping session that starts and ends in ChatGPT or Perplexity is one Amazon does not see, cannot attribute, and cannot monetize.

Amazon’s counter is Rufus, its AI shopping assistant (rebranded “Alexa for Shopping” in May 2026). Rufus reached 300 million customers in 2025. On Black Friday 2025, Rufus-assisted sessions converted at 3.5 times the rate of sessions without it, per Amazon’s own figures. Amazon is building toward a world where AI- driven shopping discovery runs through Amazon’s own AI, not a third party’s.

The transaction data that makes Rufus credible is the same data Amazon is protecting by blocking Gmail’s inbox parsing and every AI company’s web crawler. It is one asset, defended on two fronts.

The Pattern Across a Decade

This closure follows a pattern that has run consistently for ten years.

In 2013, Google removed the search referrer from organic traffic, ending keyword- level intent data for publishers. In 2017, Apple’s Intelligent Tracking Prevention cut off cross-site attribution. In 2021, Apple’s Mail Privacy Protection made email open rates unreliable. Each move was framed around user benefit. Each also protected the closing platform’s data advantage against competitors.

The brands that handled each prior closure better than their peers shared one thing: structured data on their own domain, in systems they controlled, before the pipe went dark. The brands that had outsourced their data visibility entirely had to start over each time.

The Technical Layer That Matters Right Now

There is a specific technical consequence of the AI crawler era that is concrete enough to act on.

A December 2024 analysis by Vercel of more than 500 million AI crawler requests found that none of the major AI crawlers render JavaScript. ChatGPT fetches JavaScript files in about 11% of requests but does not execute them. Anthropic’s ClaudeBot downloads JavaScript in roughly 24% of requests and never executes it. Only Gemini, which runs on Google’s existing crawler infrastructure, renders JavaScript.

If a brand’s product detail page loads reviews via a JavaScript call that fires after the initial HTML is delivered, those reviews are absent from the page an AI agent reads, regardless of how many have been collected.

Server-rendered content is what crawlers and AI agents actually access: the HTML that exists at page load, before any JavaScript runs.

A Deloitte survey cited in Digiday found 56% of consumers plan to use AI chatbots to compare prices, and 47% plan to use AI to summarize reviews before purchasing. AI-referred visitors already spend 48% more time on site and view 13% more pages per visit than non-AI traffic.

When a shopper asks ChatGPT what outdoor grill to buy, the answer comes from what the model can actually read. A brand with server-rendered reviews, AI- generated summaries baked into the page HTML, and structured product data on its own domain has a surface that crawlers can work with. A brand without it does not, regardless of how many reviews it has collected on Amazon or elsewhere.

Two Checks Worth Running Before End of Week

Check 1

Ask an AI about your hero SKU:

Open ChatGPT, Perplexity, or Gemini and search for your product by name, or by a real shopper question in your category. See whether your site surfaces and whether review content appears. Then inspect your PDP’s source HTML at page load, before any JavaScript runs. If review content is not in the raw output, it is invisible to AI crawlers. Yotpo Discover is also a great place to start :)

Check 2

Run the dependency audit:

If Amazon’s seller analytics went dark the way its email data just did for Google, what would your team still have? What do you know about your buyers that does not require a platform’s continued cooperation? The answer to that question is a map of your actual data exposure.

Amazon is not acting unusually for a platform at this stage of competition. Every major player runs this playbook when the stakes get high enough. The email change and the AI crawler blocks are the latest moves in a conflict that has been running for years.

The brands building their own crawlable, structured, server-rendered data are building something that survives these moves.

Hope this was helpful. If it was, feel free to share it with a friend.

Tomer

Track your visibility for free here: Commerce GPT Visibility Tool