Same AI, different doors: why Amazon blocked Muse while Shopify offers a connection

The same AI arrives to shop. One store blocks it; another offers a way to connect.

Amazon has reportedly blocked Meta's AI agent, Muse. Meanwhile, Shop's official help lists Muse as an example of a supported AI platform.[1][3]

The difference involves more than confidence in the AI's capabilities. Who is visiting, has the store agreed to participate, and who approves the purchase? These are distinct questions for an ecommerce team.

How Amazon and Shopify have responded to the AI agent Muse

Amazon has blocked Muse, while Shopify's Shop explains how to connect with supported agents. This does not establish that Amazon rejects every AI service or that every Shopify store permits every automated transaction.[1][3]

GeekWire reported the Amazon block on September 20, 2026. Shop's official help, checked on September 26, describes connections with supported services such as Muse.

The help page also says available features depend on the agent, store, payment method and transaction. Its update date was not visible, so it cannot establish a particular day when every store became available.[3]

Source checked	What it establishes	What it does not establish
Amazon reporting	A reported block on Muse	A permanent ban on every AI service
Shop help	Instructions for supported connections	Automatic purchases across every store and transaction

For an ecommerce team, the practical lesson is that “Should we welcome AI?” is only the first question. Product search and order confirmation require different decisions.

Amazon's concern about agent identification, and Meta's explanation

Amazon raised concerns about Muse identifying itself openly and respecting a service provider's decision to participate. Concerns about credentials were also reported as Amazon's position.[1]

A company's concern should not be presented as a confirmed incident. This reporting does not establish that passwords were leaked.

In its September 8 announcement, Meta described secure credential storage that keeps passwords and payment details out of Muse's view. It also described user confirmation before a purchase.[2]

An AI provider's security explanation and a store's agreement to accept its visits are separate matters.

For example, an ecommerce team might decide whether an approved integration and an unconnected automated visitor should have the same access. This is our editorial interpretation of the issue, not a claim about either company's internal implementation.

Under what conditions does Shopify support AI purchases?

Shop's help describes connecting an AI platform, searching for products and approving purchases. The ability to connect should not be read as unrestricted permission to pay.[3]

Separate connecting to Shop from approving a purchase

A connection enables an AI service to use Shop features. Purchase approval concerns a specific order: its products, quantities and total price.

The help describes reviewing and approving purchases by default. Available settings and approval arrangements vary with the service and other conditions, so implementation teams should inspect the actual flow for their intended transaction.[3]

Meta's September 8 announcement described Shop Pay support as forthcoming. Keeping that dated announcement separate from the subsequently checked help page avoids confusing a planned integration with its later availability.[2][3]

Our article on Muse and Stripe Link provides more context on the payment connection.

What your store should decide about identity and permitted actions

Separate visitor verification, permitted operations and order approval. The following is an editorial worksheet for discussions with your ecommerce and development teams.

Decision	Concrete question
Visitor verification	Which AI service can we verify as the visitor?
Permitted operations	Do we allow search only, cart additions, or orders too?
Order approval	Who confirms the products, quantities and total?
Records	Can we record which action succeeded and where it stopped?

Web Bot Auth concerns identity; WebMCP describes operations

Web Bot Auth uses mechanisms including signatures in HTTP requests to verify a bot's identity. A valid signature does not, by itself, grant permission to buy from a store.[4]

WebMCP lets a site expose available operations as tools for AI. Describing a product-search operation serves a different purpose from verifying the visitor's identity.[5]

We have not established that Muse uses Web Bot Auth or WebMCP in this case. These mechanisms help explain the separate problems raised by the news.

Start by naming one operation you want an agent to perform, such as product search. Then specify who can use it and where human approval becomes necessary. That turns a broad goal of “supporting AI” into a decision your team can implement and review.