🔍 Read the full analysis: How AI Shopping Agents Could Change Retailer-Customer Ties 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
AI shopping agents are beginning to search for products and, in some cases, complete purchases, but retailers are taking different approaches to letting them in. Meta says more than 10 million people in the United States and Canada set up its Muse agent in about six weeks; those figures are company-reported, and broader adoption and checkout activity remain uncertain.
AI shopping agents are prompting retailers to decide whether to admit automated buyers or direct shoppers to their own tools. Walmart announced an Agent Gateway and spending controls on October 21, while Amazon has blocked Meta’s Muse agent from accessing its site. The split matters because agents could influence which products people see, where they buy and which company retains the customer relationship, a shift that could feature in what changes in AI by 2027.
Meta says more than 10 million people in the United States and Canada set up Muse within roughly six weeks of its September launch. On Meta’s October 21 earnings call, the company said shopping was the agent’s most common use and the average purchase exceeded $200. These are Meta’s own figures; the company has not established how often users complete purchases or how the figures compare with conventional online shopping.
A report by The Verge illustrated the limits of an agent’s reach: when Muse searched for the lowest per-piece price for granola bars, it named Sam’s Club, Walmart and Costco, but said it could not check Amazon because access had been blocked. Amazon said Muse had not identified itself as an agent and raised concerns about customer credentials. Meta has said Muse cannot see users’ passwords or payment details. The competing accounts leave a dispute over access and security, rather than a settled finding about what data the agent can access—a question closely tied to digital trust.
Walmart’s October announcement offers a different route, much like the everyday errands AI agents could handle. Its Agent Gateway is designed to let approved agents search inventory and make purchases through a structured connection; its Agent Profiles let customers set permissions and spending limits. Walmart has not disclosed broad adoption or performance results. Separately, Meta has announced retail connections with several merchants, while Shopify has added tools intended to let browser agents use structured checkout functions rather than imitate human clicks.
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
Who Controls the Shopping Relationship
The commercial stakes extend beyond whether an agent can find a low price. If shoppers delegate product research and buying, the agent may become the main interface between a customer and a retailer. That could shift control over product discovery, customer data and purchase decisions toward the company operating the agent, while retailers may lose direct contact with shoppers or gain a new sales channel.
Advertising and referrals are part of the dispute. The source report cites GeekWire’s figure that Amazon made more than $68 billion from advertising last year, underscoring why control of shopping discovery can matter to retailers. Meta chief executive Mark Zuckerberg said the company expects to earn a small fee from transactions. Those figures and statements describe potential business incentives, not evidence that agents have already displaced existing advertising or referral revenue.
For shoppers, convenience also brings practical questions: who approves a purchase, what happens if an agent orders the wrong item, and which company’s return rules apply? Walmart’s spending controls address some of those concerns, but the protections and responsibilities may differ across services. Consumers should check permissions and purchase limits before delegating transactions.
Products Worth Considering
Apple iPad Air 11-inch (M2), 512GB, Wi-Fi, Space Gray
Amazon Kindle Paperwhite Signature Edition (12th Generation, 2024), 32 GB, Metallic Black
LG UltraGear 27-Inch QHD OLED Gaming Monitor, 240Hz, 0.03ms, G-SYNC Compatible, FreeSync Premium Pro, HDMI 2.1
As an affiliate, we earn on qualifying purchases.
From Browser Bots to Retail Gateways
Muse launched on September 8, according to the source report, as an agent able to browse, fill out forms and act for a user. It operates in a virtual machine in Meta’s cloud and uses Link by Stripe with a one-time-use card, according to the report. Amazon restricted its access on September 20, saying the agent violated its Conditions of Use. Meta and Amazon gave different accounts of the agent’s access and identification.
Retailers and technology firms are now trying several approaches. An agent can browse a normal website as a person would, connect to a retailer through an approved interface, or use shared protocols being developed by industry groups. Walmart’s gateway and Shopify’s checkout tools represent structured access; protocols such as the Personal Agent Protocol and Google’s Universal Commerce Protocol aim to define how agents and merchants communicate. The source describes the Personal Agent Protocol as a draft in October, while UCP had been published.
Early activity suggests agents are used more for research than payment. HUMAN Security’s September data, as cited in the source, found 81% of agent requests went to product and search pages and 2.4% to checkout and payment. Those figures refer to requests in that dataset, not the share of all shoppers or purchases. A separate example of the limits of agent-led checkout came in March 2026, when OpenAI said it would let merchants use their own checkout experiences after its initial ChatGPT Instant Checkout offered less flexibility than intended.
“Continued access by an unauthorized AI agent violates Amazon’s Conditions of Use, to which our customers have agreed.”
— Amazon, in a message displayed to Muse users, as reported by The Verge
Adoption, Access and Accountability
It remains unclear how many agent users complete purchases, how frequently they return, or whether the average purchase reported by Meta represents typical behavior. The company has not provided an independent comparison or a detailed methodology for its adoption figures. The HUMAN Security figures describe traffic requests in a particular dataset and cannot show how much of retail sales agents generate.
Access is also unsettled. Some retailers are opening structured routes, while Amazon and other firms have restricted or challenged outside agents. The source report describes cases where Muse encountered a Target bot check inconsistently and users had difficulties with travel services; those examples do not establish a general pattern across retailers. It is also not yet clear how disputes over mistaken orders, refunds, privacy or payment authorization will be handled when an agent acts across multiple companies.
Several industry protocols are at different stages of development, and participation does not guarantee that retailers will adopt them or that different agents will work across the same services. The rules governing consent, identity, data access and transaction fees remain in flux.
Retailers Test Rules and Integrations
The next signal will be whether Walmart’s gateway moves from announcement to widely used shopping connections, and whether Agent Profiles give customers meaningful control over individual purchases. Retailers and agent companies are also expected to keep testing integrations and shared protocols; their adoption, reliability and treatment of customer data will determine whether agents can move beyond product research.
For now, shoppers can expect different rules from different merchants. Before enabling an agent, users can review what it may access, require approval before each purchase, set a spending cap and check which payment method and return policy apply. The central question is not only whether an agent can buy something, but who is accountable when it does.
Key Questions
What is an AI shopping agent?
It is software that can search for products and, depending on its permissions and retailer access, take steps toward making a purchase for a user. Some agents browse ordinary websites; others connect through retailer-approved tools.
Why has Amazon blocked Meta’s Muse?
Amazon said Muse was an unauthorized agent that did not identify itself and raised concerns about customer credentials. Meta has disputed the password and payment-access concern, saying Muse cannot see those details. The two companies’ accounts are not the same.
How does Walmart’s Agent Gateway work?
Walmart says its gateway lets approved agents connect to store services, including inventory search and purchasing, through a structured interface. Its Agent Profiles are intended to let customers set permissions and spending limits. Wider availability and results were not established in the source material.
Are shopping agents already completing many purchases?
The available figures suggest much agent activity is focused on product and search pages rather than checkout. HUMAN Security reported that 2.4% of agent requests in its September data went to checkout and payment; that dataset does not represent all shoppers or retail transactions.
What should shoppers check before using an agent?
Review what the agent can access, set a spending limit, require approval for purchases if available, and confirm which payment method and retailer return policy apply. Controls vary by agent and merchant.
Source: ThorstenMeyerAI.com
