🔍 Read the full analysis: Who Owns The Customer When AI Shops Across Retailers? on ThorstenMeyerAI.com
Get business pricing on tech for your team
- Business-only prices and quantity discounts
- Tax-exempt purchasing
- Multiple users, one account, clear invoices
TL;DR
Meta says more than 10 million people in the United States and Canada set up its Muse agent within about six weeks, with shopping its most common use. Walmart is building access for outside agents, while Amazon has blocked Muse, bringing a dispute over access, customer relationships and transaction credit into focus.
Meta says more than 10 million people in the United States and Canada set up its Muse shopping agent in roughly six weeks, as Walmart opens a route for outside agents and Amazon blocks Muse from its site. The contrasting moves sharpen a commercial question for retailers, publishers and shoppers: who controls the customer relationship when software searches across stores and may place an order?
Meta shared its Muse figures during its October 21 third-quarter call. The company said shopping was the agent’s most common use and that the average purchase was worth more than $200. These are Meta-reported figures; the source material does not provide an independent audit or explain how the average purchase was calculated.
Walmart announced an Agent Gateway and Agent Profiles, which are intended to let approved agents access shopping services with customer-set spending limits. Amazon has taken the opposite approach with Muse. After blocking the agent in September, Amazon said it had not been told Muse would shop on its site, did not identify itself as an agent and appeared to capture and store customer credentials. Meta disputed the credentials claim, saying Muse cannot see people’s passwords or payment methods.
The immediate issue is not simply whether an agent can find a product. It is whether it can access inventory, carry a shopper’s permission through checkout and make clear which company handles payment and returns. A September analysis from HUMAN Security cited in the source material found 81% of observed agent requests went to product and search pages, while 2.4% went to checkout and payment. Those figures describe that dataset, not all agent shopping.
Agentic shopping
Three ways an AI agent buys for you
Every fight between shops and AI agents comes down to which of these three routes the agent uses, and whether the shop agreed to it.
The browser, like a human
The agent opens pages built for people, signs in with your login and clicks through checkout. Works almost anywhere, which is why shops object: they can't tell the agent from you.
Examples: Meta's Muse (from a virtual machine in Meta's cloud), Perplexity's Comet (from your own computer)
The shop's own agent door
Approved agents search and buy through a structured interface instead of clicking through pages. The shop sets the rules.
Examples: Walmart Agent Gateway (MCP, beta), Shopify MCP server and WebMCP checkout tools
Shared protocols
Common rules for who the agent is, what you allowed, and how the checkout and payment are confirmed.
Examples: Personal Agent Protocol (draft), Universal Commerce Protocol, Agentic Commerce Protocol
The protocols competing to be the standard
| Protocol | Backers | Covers | Status |
|---|---|---|---|
| Personal Agent Protocol (Poppy) | Sierra, Meta, Walmart, Shopify, Stripe + 35 design partners | Agent identity, customer permission, one session across site, API and company agent | Draft 0.1 (9 Oct 2026) |
| Universal Commerce Protocol (UCP) | Google, Shopify; site also lists Etsy, Wayfair, Target, Walmart, Amazon, Microsoft | Discovery, cart, checkout, orders; payment mandates via AP2 | Published |
| Agentic Commerce Protocol (ACP) | OpenAI, Stripe | Checkout between agent and merchant | Open source; ChatGPT's own checkout scaled back in March 2026 |
What a careful agent purchase looks like
- You give the task and a limit, for example a $50 cap.
- The agent identifies itself to the shop and gets only the access you approved.
- It prepares the cart and shows you an approval card.
- Payment goes through a token or one-time card, so the agent never sees your card.
- The shop’s normal returns apply if the agent got it wrong.
Sources: Meta (8 Sep 2026), Walmart (21 Oct 2026), TechCrunch on Shopify (28 Sep 2026), Sierra (9 Oct 2026), ucp.dev, agenticcommerce.dev, OpenAI (24 Mar 2026). Purchase steps: Walmart’s agent terms and Meta’s Muse description.
Powered by Thorsten Meyer AI
Retailers Contest the Agent Relationship
When an agent selects products across multiple stores, it can become the shopper’s main interface. The retailer may still fulfill the order, but the agent can shape which products are considered and where a purchase is made. That raises commercial questions about customer access, advertising and transaction fees, as well as practical questions about who is responsible when an order goes wrong.
The source material says Amazon earned more than $68 billion from advertising last year, citing GeekWire, and reports that Meta expects to take a small fee from transactions. Those figures and statements point to competing business interests, but they do not establish how much revenue any company will ultimately earn from agent-led shopping. Retailers may favor direct relationships and control over checkout; agent providers may seek payment for referrals or completed purchases.
Publishers and affiliate businesses also have a stake. Amazon Associates commissions are tied to customers clicking tracked links, according to the source material. If an agent reads a review and buys through another route, a publisher may not receive the click attribution that normally supports a commission. The scale of that effect is not established, but the shift could change how product recommendations are funded.
For shoppers, a smoother purchase does not remove the need to know who is charging the card, what spending authority was granted and which return policy applies. Walmart’s profile limits are one approach to permission. Meta’s Muse description includes payment through Link by Stripe using a one-time-use card, but the source does not establish that this arrangement works identically across every retailer.
Products Worth Considering
LG UltraGear 27-Inch QHD OLED Gaming Monitor, 240Hz, 0.03ms, G-SYNC Compatible, FreeSync Premium Pro, HDMI 2.1
Bose QuietComfort Ultra Earbuds (2nd Gen), Deep Plum
Bose TV Speaker (2nd Gen) Compact Soundbar with Bluetooth and HDMI Connectivity
As an affiliate, we earn on qualifying purchases.
From Browser Bots to Retail Gateways
The conflict developed quickly. Meta launched Muse on September 8, describing an agent that could use a browser, fill out forms and act for a user. It runs in a dedicated virtual machine in Meta’s cloud. Amazon blocked it on September 20. Three days later, Meta announced retail connectors with Best Buy, Gap, Sephora, Walmart and Wayfair, among others.
Other companies are trying to make agent access more structured. Shopify introduced tools for browser agents to complete checkout, while Walmart’s October announcement offered an approved gateway. A separate effort, the Personal Agent Protocol, or Poppy, was published as a draft on October 9 by Sierra and Meta, with Walmart, Shopify and Stripe among its backers. The source material also describes other commerce protocols, including Google- and Shopify-backed UCP and OpenAI and Stripe’s ACP.
These systems reflect three broad routes: agents can browse websites like people, connect through a retailer’s own interface, or use shared protocols. Browser-based access can work on sites not designed for agents, but retailers may have difficulty distinguishing automated activity from a shopper’s actions. Structured interfaces can provide clearer permissions, but depend on stores choosing to support them. Protocols are still developing, and their presence does not mean that every retailer or payment provider has adopted them.
The source material notes that OpenAI stepped back in March 2026 from its initial Instant Checkout approach, saying the first version lacked the flexibility it wanted and that it would let merchants use their own checkout experiences while focusing on product discovery. That example shows that discovery and checkout are separate challenges: an agent may recommend an item without completing the transaction itself.
Access, Attribution and Liability Remain Open
It is not yet clear how many Muse users have completed purchases, how often their agents successfully reach checkout, or how Meta’s reported average purchase value was measured. The adoption figures came from Meta, and the source material provides no independent verification. The HUMAN Security percentages cover its cited observations; they should not be treated as a market-wide rate.
The longer-term rules are also unsettled. The source describes Poppy as a draft and says payment details were left to extensions, while other protocols have different backers and functions. It does not establish which standards will gain broad retailer support. Nor does it spell out how responsibility will be divided among an agent provider, retailer and payment company if an agent exceeds a user’s permission, orders the wrong product or mishandles personal information.
Amazon’s stated reasons for blocking Muse and Meta’s response remain competing company accounts. The source material does not provide a technical audit resolving whether Muse accessed or stored credentials. Retailer policies, agent capabilities and court or regulatory developments may change as these systems are tested.
Watch Checkout and Permission Controls
The next signal will be whether Walmart’s Agent Gateway and profiles move from announcement to widely available shopping features, and which outside agents retailers approve. For Meta, the key questions are whether Muse can complete more purchases through retail connectors and how users will manage authorization, payment and returns across stores.
Retailers, payment providers and protocol developers are also likely to keep refining how an agent identifies itself and proves what the customer allowed it to do. Shoppers considering these tools should check purchase approvals, spending limits and card controls before granting access, and confirm which merchant’s policies govern delivery and returns. The reach of agent shopping—and who receives credit or revenue when it happens—remains in development.
Key Questions
What is the main development in AI shopping?
Meta reported rapid setup of its Muse agent, while Walmart announced a gateway for approved agents and Amazon continued to block Muse. The moves show different approaches to outside software shopping on retailer sites.
Why did Amazon block Meta’s Muse?
Amazon said Muse was an unauthorized agent, had not identified itself and appeared to capture and store customer credentials. Meta disputed the credentials concern, saying Muse cannot see users’ passwords or payment methods. The source material does not include an independent technical review resolving the dispute.
Can AI shopping agents complete a purchase?
Some agent and retailer tools are designed to support checkout, but the source material indicates that many observed agent requests were for product and search pages rather than payment. Research and purchase completion are not the same step, and availability depends on the agent and retailer.
What should shoppers check before using an agent?
Check whether purchases require approval, set a spending limit and understand how payment is handled. Confirm which retailer’s delivery, cancellation and return rules apply before relying on an agent to place an order.
Source: ThorstenMeyerAI.com
