Amazon's Wall

The decision

On the night of Sunday, September 20, Amazon started blocking Muse, Meta’s AI agent, from its store. Muse can browse and buy on a user’s behalf. Amazon had asked Meta to pull the agent from its site, and Meta declined.

Amazon’s reasons were specific. It said Muse “does not identify itself as it browses” and “appears to capture and store customer credentials.” Meta said Muse “has no visibility into people’s passwords or payment methods.” Amazon’s broader line was that shopping agents should “operate openly and respect service provider decisions about whether or not to participate.”

Muse isn’t alone. Amazon also blocks shopping agents from OpenAI’s ChatGPT, Google’s Gemini and Perplexity’s Comet browser.

The next morning, Shopify went the other way. CEO Tobi Lütke announced the company was “partnering deeply with Muse to enable agentic checkout with Shop Pay on all Shopify stores.” Mark Zuckerberg posted that Meta was “teaming up with Shopify to make shopping and checkout easier in Muse.” Shopify’s stock rose 7.3% that day and about 15% over 2 days, after falling around 21% so far in 2026.

Same agent, same week, opposite answers. That is the decision.

The strategic why

Start with Amazon’s ad business. It made $19.8 billion in Q2, up 26% from a year earlier. Almost all of it depends on one thing: a human looking at an Amazon page. Sponsored products sit at the top of search results because people scroll past them. Brands pay for that placement because people click.

An agent doesn’t scroll. It doesn’t see a sponsored slot, and it doesn’t care which brand paid for the top spot. It reads the product data, compares prices, possibly across other stores, and buys the best option. To an agent, Amazon is a warehouse with a checkout. An agent shopping on Amazon would collect the convenience of Prime and skip the most profitable part of the visit.

That’s why the block is rational. Amazon isn’t protecting its customers from Meta. It’s protecting a $20 billion-a-quarter toll booth on human attention.

There’s a second reason: Amazon wants to be the agent. In May, it moved its shopping assistant into Alexa as a shopping agent, and its “Buy for Me” feature can buy from other retailers’ sites on a customer’s behalf. Merchants complained that Buy for Me listed their products without asking first, and the only way out was an email opt-out. Amazon’s rules for agents are simple: other people’s agents need permission, and Amazon’s own agent asks for forgiveness.

Now look at Shopify. It has no ad business to protect and no consumer front door to lose. It earns money when something sells. Its president, Harley Finkelstein, made the economics plain: “agentic transactions carry the exact economics as an online store transaction.” Shopify collects about 2.4 cents of every dollar its merchants sell, and it doesn’t matter to Shopify whether a human or an agent pressed the button.

Shopify’s millions of independent merchants are also the stores that can’t outspend big brands on Amazon ads. An agent that ignores ad placement and judges products on price and fit is the best thing that could happen to them. Shop Pay had processed more than $400 billion in lifetime sales by June. Now it’s the checkout button inside the most downloaded free iPhone app in the US, which Muse became within 10 days of launch.

Amazon owns the store, so agents threaten it. Shopify owns the checkout, so agents feed it.

What it defends against

For Amazon: becoming a supplier to someone else’s front door. Amazon spent 30 years teaching customers to start their shopping on Amazon. If customers start on Muse instead, Amazon becomes one store among many in an agent’s comparison. The block keeps the first click on Amazon for as long as Amazon can hold it.

For Amazon: losing the legal fight. Amazon tried the courts first. In August, the Ninth Circuit threw out its injunction against Perplexity’s Comet. The court said it’s the user, not the AI company, who “accesses” Amazon under the federal hacking law, which leaves Amazon’s hacking claims weak. But the court left Amazon’s terms of service and technical blocks intact. Blocking Muse is Amazon using the one weapon the court left it.

For Shopify: being invisible. Shopify’s merchants have always had to buy their customers through Google, Meta and Amazon ads. If agents become the way people shop, being the default checkout inside them is worth more than any ad slot. Shopify got there in one day.

The risk

Amazon’s risk is a wall around an emptying castle. Blocking works if agents stay a niche and Amazon’s own agent catches up. It fails if customers decide they like asking Muse or ChatGPT to shop for them more than they like searching Amazon. Then the block doesn’t keep customers in. It keeps Amazon out of the place they’ve gone. The Buy for Me double standard also gives regulators an easy story to tell.

That’s why we’re moving Amazon’s disruption resilience from 8 back to 7, taking back the point AWS earned it two weeks ago. AWS is set up to profit from agents. The retail and ad business, so far, is set up to keep them out. Amazon moves from 3.94 to 3.88. It’s still Deep.

Shopify’s risk is its new partner. Meta now decides which Shopify stores Muse recommends, and in what order. That’s the same position Meta holds over Shopify merchants’ ad budgets today. Meta’s stock dipped about 0.5% on the deal because investors noticed Meta handed the checkout to someone else. Expect Meta to want it back.

Meta’s risk is the one our Muse article flagged. Muse found a way into commerce, but the door belongs to Shopify. Meta’s distribution control stays at 4. It still rents the road.

The agents are coming either way. Amazon is betting it can make them come through its own door.