Rufus Isn't Just Recommending Products Anymore. It's Buying Them.

Angie Chua runs Bobo Design Studio. She built her brand on Shopify and made a deliberate choice not to sell through Amazon. Then orders started arriving from Amazon customers.

Amazon's AI had enrolled her in "Buy for Me" — a feature that lets Rufus complete purchases from brand websites on behalf of shoppers — without asking. She didn't opt in. She wasn't notified. She found out when the orders showed up. As she put it: "We were forced to be dropshippers on a platform that we have made a conscious decision not to be part of."

This is where Amazon Rufus crossed into new territory. It's not a recommendation engine anymore. It's an agent that can complete a transaction.

How It Works

When a shopper does a branded search on Amazon and the product isn't in Amazon's catalog, Rufus now shows two options. A "Buy for Me" button lets Amazon's AI complete the purchase using the shopper's saved payment and shipping information — the shopper never leaves the Amazon app. A "Shop Direct" button sends the shopper to the brand's website to complete the purchase themselves.

Either way, the brand handles fulfillment, returns, and customer service. In the "Buy for Me" case, Amazon handles the transaction interface and captures the data.

The feature launched in April 2025 with 65,000 products. It now covers more than 500,000. Amazon projects it will facilitate over $10 billion in annual sales. The default was opt-out, not opt-in. That controversy broke publicly in January 2026 — and Amazon's default hasn't changed. Brands on Shopify and other platforms were pulled in without being notified. The Bobo Design Studio situation wasn't an edge case — it was the model.

What Changes When the AI Buys

A recommendation and a purchase are different things. This distinction matters more than it might sound.

When Rufus recommends a product, the shopper still decides. They see the suggestion, evaluate it, and choose to buy or not. The brand's product page, pricing, and whatever else catches the shopper's attention all have a chance to influence the outcome. There's a gap between the recommendation and the purchase, and that gap is where brand experience works.

When Rufus completes a purchase, that gap closes. The AI decided and acted. The shopper's role was authorizing the agent to work on their behalf — which they did when they set up the feature, not at the moment of this particular transaction.

The only leverage you have when AI is completing the purchase is how you're represented in the recommendation that precedes it. If Rufus doesn't recommend your product clearly and accurately, the transaction doesn't happen. There's no checkout page to optimize, no abandon cart email to send, no second chance.

What Amazon Captures From These Transactions

Amazon says it uses encrypted customer data to complete "Buy for Me" transactions but "cannot see unrelated or past orders made on the brand's site." That's a narrow claim.

What Amazon does see: the product, the price you charged, whether the transaction completed, and fulfillment data. Across 500,000 products from brand websites that chose not to work with Amazon, that's a meaningful amount of intelligence — pricing data from brands outside their ecosystem, transaction outcomes, and whether their AI is good at converting the shoppers it routes.

This isn't Amazon doing something illegal. But it's worth understanding what the data picture looks like at scale.

What This Means If You're Selling on Amazon

For brand teams running Amazon operations, "Buy for Me" is primarily about brands that aren't in Amazon's catalog. If your products are listed on Amazon, a shopper searching your brand name will typically see your Amazon listings, not the "Buy for Me" flow.

But two situations connect this to your world directly.

First, if your brand has both Amazon listings and a DTC site, your website may be enrolled in "Buy for Me" even if your Amazon listings are active. Shoppers looking for a specific SKU you sell on both channels could trigger the "Buy for Me" experience for your direct site. That's a customer relationship Amazon is now intermediating.

Second, your competitors who don't sell on Amazon can now appear in Amazon search via "Buy for Me" — and Rufus can route customers to their websites. The question of what Amazon Rufus recommends is now broader than Amazon's catalog. Brand teams who only track their visibility within Amazon's marketplace are watching a partial picture.

What to Do About It

Check your enrollment status. Open the Amazon Shopping app and search your brand name. If you see "Shop brand sites directly" or a "Buy for Me" button associated with your brand, you're in the catalog. That's your starting point for deciding whether you want to stay in or opt out.

If you don't want this: Amazon does have an opt-out process through Brand Registry and Seller Support. You have to actively find it — it isn't surfaced proactively.

If you're staying in (or already on Amazon): The quality of the "Buy for Me" experience depends entirely on what Rufus knows about your brand. The same entity signals and listing accuracy covered in recent posts affect how Rufus handles your products in an agentic purchase flow. Weak product information creates mistakes — wrong SKUs, mismatched variants, inaccurate availability. Those mistakes are brand problems you don't see until a customer complains.

Track what Rufus recommends in your category's "Buy for Me" results. Which competitors are showing up? Which brand websites is Rufus routing Amazon shoppers to? That's competitive intelligence most Amazon analytics tools don't surface at all.

The Line Amazon Just Crossed

The shift from Amazon Rufus as a recommendation tool to Amazon Rufus as a purchasing agent isn't just a product feature update. It's a different relationship between Amazon and every brand with a web presence — regardless of whether that brand ever chose to work with Amazon.

Rufus processing 274 million queries per day is a visibility problem. Rufus completing purchases on behalf of 300 million shoppers is a brand control problem. The optimization work looks the same — make sure Rufus represents your products accurately. But the stakes are different. When AI is completing the purchase, getting your information wrong doesn't mean a lost click. It means a wrong transaction.

Most brand teams still don't have a system for tracking what Amazon Rufus says about their products, let alone how it routes purchase decisions. That's what AgentBuy monitors: what Rufus actually says about your brand and category, including how it handles branded searches that lead to "Buy for Me" and "Shop Direct" recommendations.

The default was opt-out. If you didn't know you had a choice, now you do.