What Is Amazon Buy for Me? How It Compares to Universal Checkout
Arjun Bhargava
Co-founder and CEO @ Rye
9 minutes read
Amazon's Buy for Me proves commerce won't be solved until it's universal — but it locks that universality inside a walled garden. Here's how the alternatives compare.
TL;DR / Key Takeaways
Amazon's Buy for Me uses agentic AI to purchase from third-party brand sites on customers' behalf — no merchant integration, no merchant consent. As of March 2026, it covers tens of millions of products from 400,000+ merchants.
Buy for Me is locked inside Amazon's app. There's no API, no developer access, and no way to integrate it into your own product. If you're building an AI agent, it's not an option.
Open protocols like ACP and UCP take the consent-first path — but only ~30 merchants ever went live on ACP before ChatGPT killed its native checkout in March 2026. UCP is growing, but adoption is still early.
Universal Checkout APIs work on any merchant site via product URL. No protocol adoption, no merchant integration, no platform lock-in.
Amazon proves what we've seen building Rye: commerce won't be solved until it's universal. Customers would rather have an AI agent fumble through a merchant's checkout than leave the app they're already in. The question isn't whether every product needs to be purchasable everywhere — it's who builds that infrastructure and on what terms.
How Amazon Buy for Me Actually Works
Amazon launched a feature that sends AI agents to scrape 400,000 merchant websites and complete purchases on their behalf. Then Amazon sued Perplexity for doing the same thing to Amazon.
That's the short version of Buy for Me — a beta feature in the Amazon Shopping app that uses agentic AI to purchase products from third-party brand sites on the customer's behalf. The customer never leaves Amazon. The brand often doesn't know Amazon is involved until the order arrives.
Buy for Me launched in April 2025. By March 2026, Amazon had expanded it significantly — Shop Direct now indexes over 100 million products from more than 400,000 merchants, with tens of millions purchasable through the AI agent. Amazon added merchant feed integrations through Feedonomics, Salsify, and CEDCommerce, giving brands the option to sync catalogs voluntarily. But participation isn't required. Amazon scrapes public product data by default.
Under the hood, the system runs on Amazon Bedrock with Nova and Anthropic's Claude models. Amazon generates a unique email address per order so brands can't directly contact the customer. The brand handles fulfillment and returns, but the customer relationship stays with Amazon — no email capture, no loyalty enrollment, no cross-sell opportunity for the merchant.
But the most interesting question about Buy for Me isn't the mechanics. It's why Amazon needs it at all.
The Real Reason Amazon Built Buy for Me
Amazon already has hundreds of millions of products — more selection than any retailer in history. So why build an AI agent to buy from other brands' websites?
Because even Amazon can't sell everything. Third-party sellers account for more than 60% of all units sold on Amazon. And there are entire categories — niche DTC brands, luxury goods, regional merchants, products that have never been listed on Amazon — that Amazon's marketplace can't reach no matter how large it grows.
Buy for Me is Amazon's answer to the long tail. Rather than onboard every merchant onto its marketplace, Amazon built an AI agent that goes to the merchant's site and buys on the customer's behalf. The customer stays in the Amazon app. Amazon captures the shopping data. The merchant gets the order.
Here's what makes this telling: the Buy for Me experience is objectively worse than buying directly from the merchant. Order and tracking information doesn't sync properly. Customers have to contact brands separately for support. There's no loyalty enrollment, no personalized offers, no relationship with the brand. And yet customers are using it — Buy for Me inventory has grown from 65,000 products at launch to over 500,000, and Amazon keeps expanding it.
That's the signal. Customers would rather have an AI agent fumble through a merchant's checkout than leave the app they're already in. Convenience beats experience. 43% of consumers will spend a maximum of three minutes on checkout before abandoning entirely. The friction of leaving one app, finding a merchant site, creating an account, and entering payment details is enough to kill the sale.
Amazon sees the same thing we see at Rye: commerce won't be solved until it's universal. Every product on the internet needs to be purchasable from wherever the customer already is. Amazon is building that inside its walled garden. The question is whether the rest of the ecosystem gets the same infrastructure — or whether Amazon is the only one that can offer it.
How Amazon Chose to Get There — and What It Costs Merchants
Amazon's approach to universal commerce is simple: scrape first, ask questions later.
Amazon's official announcement frames Buy for Me as giving "brands increased exposure and seamless conversion" and states that "brands have the choice if they want to participate." In January 2026, merchants discovered that the reality was different — Amazon had listed their products without permission, including items that were out of stock or that they'd never sold. The default is opt-in for every brand with a public website. The only way out is to contact Amazon and request removal. An IP attorney is currently seeking affected merchants for possible legal action.
Then there's the Perplexity lawsuit. In November 2025, Amazon sued Perplexity AI, alleging its Comet browser disguised AI agents as Chrome sessions to scrape Amazon and complete purchases. In March 2026, a federal judge granted Amazon a temporary injunction blocking Comet. The 9th Circuit later stayed that injunction pending appeal.
Amazon's position: when Perplexity does it to Amazon, it's unauthorized scraping that warrants an injunction. When Amazon does it to 400,000 merchant sites, it's a feature that gives brands "increased exposure."
The underlying problem Amazon identified is real — commerce needs to be universal. But Amazon's solution locks that universality inside a single app, strips merchants of their customer relationships, and applies a different standard to its own platform than it demands of others.
Two Alternatives, Two Different Bets on Universal Commerce
If Amazon is right that commerce needs to be universal, the follow-up question is: does it have to be built this way? Two other models have emerged, each trying to solve the same problem with fundamentally different tradeoffs.
Open protocols take the consent-first path. Stripe and OpenAI's Agentic Commerce Protocol requires merchants to opt in, provide structured product feeds, and maintain integration endpoints. Google's Universal Commerce Protocol follows the same model — Walmart, Target, Wayfair, and Etsy are live, with 20+ global partners endorsing it.
The advantage is clear rules. Merchants control their data and customer relationships. The disadvantage is adoption speed. OpenAI's ChatGPT Instant Checkout launched with ACP and was killed in March 2026 after near-zero conversion rates. Only ~30 Shopify merchants ever went live. UCP has more momentum, but it still requires merchant engineering resources most brands don't have.
Universal checkout APIs take the coverage-first path. The contract is simple: product URL in, confirmed order out. No merchant integration required, no protocol adoption needed. The API navigates live checkout flows, so pricing and inventory are always current rather than dependent on feed freshness.
At Rye, we built the Universal Checkout API for exactly this use case — any developer can purchase from any merchant, including Amazon via direct integration, without being locked to a single platform. The tradeoff is the same consent ambiguity as Buy for Me. The difference is that Universal Checkout makes the infrastructure available to everyone, not just Amazon. The same universality Amazon is building for its own customers becomes accessible to any agent builder.
How Buy for Me Compares
The differences become concrete when you line them up.
Amazon Buy for Me | Open Protocols (ACP/UCP) | Universal Checkout (Rye) | |
|---|---|---|---|
Merchant integration required? | No — Amazon scrapes public data | Yes — feeds, APIs, protocol adoption | No — works via product URL |
Available to third-party developers? | No — Amazon app only | Yes — via protocol spec | Yes — REST API |
Merchant coverage | 400K+ merchants (Shop Direct) | ~30 ACP merchants; Walmart, Target, Wayfair, Etsy on UCP | 15,000+ merchants |
Real-time pricing? | Yes — checks at purchase time | Depends on feed freshness | Yes — live checkout flow |
Merchant consent required? | No — opt-out only via email | Yes — explicit opt-in | No |
Customer relationship | Amazon owns | Merchant retains | Developer controls |
Commission/fees | Undisclosed (beta) | Varies by protocol | API usage pricing |
Amazon products included? | Yes (native) | No — Amazon blocked AI agents | Yes — direct integration |
Platform lock-in | Amazon app only | Protocol-specific | Platform-agnostic |
The Market Is Converging on Universal. The Question Is Architecture.
Every major player in commerce has reached the same conclusion: customers want to buy anything from anywhere, and the infrastructure to make that happen doesn't exist yet. Amazon is building it inside a walled garden. Protocols are building it through standardization. Universal Checkout is building it through automation.
Three approaches, three different bets on the same underlying problem.
Amazon's bet: the web is public. If a brand publishes product data on their website, an AI agent can act on it. The brand gets the order. They can opt out by email. The tradeoff is that the universality only works inside Amazon's app.
The protocol bet: merchants should explicitly consent. ACP and UCP require brands to build integrations, provide structured feeds, and maintain endpoints. This protects the merchant relationship but creates an adoption bottleneck that has limited real-world coverage to a handful of brands. Universal commerce through opt-in scales slowly.
The Universal Checkout bet: automate existing public checkout flows without requiring either Amazon's walled garden or protocol adoption. The merchant's website works as-is. The consent tradeoff is the same as Buy for Me, but the infrastructure is open — any developer, any agent, any merchant with a website.
Morgan Stanley projects $115 billion to $385 billion in incremental US e-commerce from agentic commerce by 2030, with roughly half of American shoppers using AI agents. The destination is universal. The only open question is how we get there.
What This Means If You're Building an Agent
Buy for Me isn't available to you. There's no API, no webhook, no way to plug it into your product. It works inside Amazon's app, for Amazon's customers, on Amazon's terms.
If your agent needs to buy Amazon products specifically, you need either a direct Amazon integration or a universal checkout API that supports Amazon — because Amazon is absent from both ACP and UCP and has actively blocked third-party AI agents from its platform.
If merchant consent is a hard requirement for your product — for legal, brand, or compliance reasons — open protocols give you that, with limited coverage today. The pragmatic production architecture is protocol where available, universal checkout for everything else.
Amazon built Buy for Me for Amazon's customers. We built universal checkout for everyone else's agents.
Frequently Asked Questions
What is Amazon Buy for Me?
Buy for Me is a feature in the Amazon Shopping app that uses agentic AI to purchase products from third-party brand websites on a customer's behalf. The customer stays inside Amazon's app while Amazon's AI navigates the brand's checkout flow, completing the transaction with encrypted customer data. The feature runs on Amazon Bedrock with Nova and Anthropic's Claude models powering the agentic capabilities. As of March 2026, Shop Direct indexes over 100 million products from 400,000+ merchants, with tens of millions purchasable through Buy for Me.
How does Amazon Buy for Me compare to ACP and UCP?
Buy for Me doesn't require merchant integration — Amazon scrapes public product data. ACP (Stripe/OpenAI) and UCP (Google) both require merchants to explicitly opt in with structured feeds and API endpoints. Buy for Me has far broader coverage but no merchant consent requirement. Protocols have consent built in but limited adoption — only ~30 merchants ever went live on ACP before OpenAI scaled back ChatGPT checkout. For a deeper comparison of how these protocols work, see our guide to agentic checkout.
Can AI agents purchase from Amazon?
Not through Amazon's own tools — Buy for Me is consumer-facing only, with no developer API. Amazon has actively blocked third-party AI agents from its platform and is absent from both ACP and UCP. Universal checkout APIs with direct Amazon integration are the primary path for developers building agents that need to purchase Amazon products.
Do merchants need to integrate for agentic checkout?
It depends on the approach. Open protocols (ACP/UCP) require merchant integration — structured product feeds, API endpoints, and ongoing maintenance. Amazon Buy for Me and universal checkout APIs do not — both work on existing merchant websites without changes. Merchants can opt out of Buy for Me by contacting Amazon directly. For a fuller breakdown of what each approach requires, see What Is Agentic Checkout?
How does Amazon Buy for Me handle customer data?
Amazon provides the customer's encrypted name, address, and payment details to complete checkout on the brand's site. A unique email is generated per order, so the brand can fulfill and communicate about that specific order but cannot add the customer to marketing lists or capture their contact information. Amazon retains the customer relationship and shopping data — the merchant has no way to remarket to or build a direct relationship with the buyer.
Start Building
Amazon built Buy for Me to make commerce universal inside its walled garden. Universal checkout makes the same infrastructure available to every developer, for every merchant, on the open web.
Product
Resources