🔍 Read the full analysis: Will AI Shopping Agents Change The Balance Between Retailers And Customers? on ThorstenMeyerAI.com
Get hardware and tech essentials delivered free with Prime
- Fast, free delivery on millions of items
- Prime Video, Amazon Music and more included
- Member-only deals all year
TL;DR
Meta says more than 10 million people in the US and Canada set up its Muse shopping agent within about six weeks, while Walmart has announced tools to let approved agents shop and Amazon has blocked Muse. The emerging contest is over customer access, purchasing data and control of checkout; how much agents will change shopping remains uncertain.
Walmart is opening a channel for AI shopping agents while Amazon has blocked Meta’s Muse, setting up a contest over which companies control the customer relationship when software searches for and buys products. Meta said on October 21 that more than 10 million people in the US and Canada had set up Muse in roughly six weeks; the figures are company-reported, and it is not yet clear how often users complete purchases through the agent.
Walmart announced Agent Profiles and a Walmart Agent Gateway on October 21. The gateway is designed to let approved agents search Walmart’s inventory and make purchases through a structured connection, rather than operating the website as a human user would. Agent profiles are intended to give customers controls such as spending limits. Walmart’s announcement points to a route in which retailers can set conditions for access instead of leaving agents to navigate ordinary webpages.
Meta’s Muse uses a browser-based approach, running in a dedicated virtual machine in Meta’s cloud. A Verge reporter testing it asked where granola bars cost least per piece. Muse 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 how it accessed customer accounts. Meta has disputed the suggestion that Muse can see people’s passwords or payment details.
Meta reported that shopping is Muse’s most common use and that its users’ average purchase is more than $200. Those are figures from Meta’s October 21 earnings call, not independently verified measurements. The available data also suggest agents currently do more product research than checkout: HUMAN Security reported in September that 81% of agent requests it measured went to product and search pages, compared with 2.4% to checkout and payment. The figures describe that company’s observed traffic, not all shopping-agent activity.
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 Agent Checkout
The dispute concerns more than whether an agent can compare prices. If customers delegate product searches and purchases to an assistant, the company controlling that assistant may influence which retailers are considered, what product information shoppers see, and where checkout happens. Retailers, meanwhile, want control over access to their inventory, customer accounts and transaction process.
That creates a commercial question about customer ownership and payment. Amazon generated more than $68 billion in advertising revenue last year, according to GeekWire, giving it an interest in the shopping activity and visibility that agents could redirect. Meta chief executive Mark Zuckerberg said the company expects to make money by taking a small fee from transactions. Those models are not yet evidence of how much revenue agents will generate, but they show why retailers and technology companies are competing to shape the route from product search to purchase.
There are practical consequences for shoppers, too. An agent may be convenient, but users need to know what it is authorized to spend, whether each purchase needs approval, and which retailer’s shipping and return rules apply. Automation does not remove purchase risk: prices, product details and return conditions still need checking, and payment protections can vary by service.
Products Worth Considering
Apple MacBook Air 15-inch Laptop with M5 Chip, 16GB Memory, 512GB SSD, and Liquid Retina Display
LG UltraGear 27-Inch QHD OLED Gaming Monitor, 240Hz, 0.03ms, G-SYNC Compatible, FreeSync Premium Pro, HDMI 2.1
Apple MacBook Pro 16.2-inch with M5 Pro, 18-Core CPU, 20-Core GPU, 48GB Memory, 1TB SSD, Nano-Texture Display, Space Black
As an affiliate, we earn on qualifying purchases.
Three Routes to Agent Shopping
The current contest involves three broad approaches. A browser agent, such as Muse, interacts with websites much as a person does, using pages and potentially signing in with the user’s credentials. That can work across many sites, but retailers may have difficulty distinguishing the software from a customer and enforcing their rules. Amazon’s Conditions of Use require outside agents to identify themselves, according to the company.
A second approach is a retailer-approved connection. Walmart’s gateway uses the Model Context Protocol (MCP) to connect agents with its services. Shopify has also introduced structured checkout tools for browser agents, allowing them to interact with a merchant’s checkout without relying only on screen-based clicks. In this model, retailers can define the functions an agent may use.
A third approach is shared standards. Meta, Sierra, Walmart, Shopify and Stripe announced the Personal Agent Protocol, also called Poppy, in October, with a draft and additional design partners. Google and Shopify have published the Universal Commerce Protocol, while OpenAI and Stripe have developed the Agentic Commerce Protocol. These projects address different parts of shopping and are at different stages; the existence of protocols does not mean retailers or agents have adopted one common system.
Recent developments show how unsettled the field is. Meta announced retail connectors involving companies including Best Buy, Gap, Sephora, Walmart and Wayfair in September. OpenAI, which launched ChatGPT Instant Checkout using its protocol in 2025, said in March 2026 that it would let merchants use their own checkout experiences while it focused on product discovery, after saying the initial version lacked the flexibility it wanted.
“Continued access by an unauthorized AI agent violates Amazon’s Conditions of Use, to which our customers have agreed.”
— Amazon, in a message shown to users accessing Amazon through Muse
Adoption and Accountability Remain Open
Meta’s user and spending figures have not been independently verified in the supplied material, and the company has not provided a clear measure of how many Muse users complete purchases. The September traffic figures from HUMAN Security offer one snapshot, but they do not establish how often agents reach checkout across the wider market or how that will change.
It is also unresolved which protocols retailers and technology companies will support at scale. Some standards are published, others remain drafts, and a retailer’s presence in a protocol’s partner list does not by itself confirm the full range of live shopping functions. The sources do not establish whether Amazon will permit any outside agents under different conditions, or how its dispute with Muse may develop.
For customers, important details may differ by agent and merchant: what purchases require approval, how spending limits work, what information is shared, and who handles disputes, cancellations and returns. Affiliate publishers also face uncertainty because commission systems tied to tracked links may not credit a sale if an agent reads a review and buys through another route. The scale of that effect is not yet established.
Retail Access and Standards Take Shape
The next test is whether Walmart’s announced tools become a usable shopping route for agents and whether other retailers follow with their own access rules. Protocol developers will also need to show how their systems handle customer permissions, identity, payment and order changes in practice. Further adoption announcements may clarify whether the market moves toward shared standards or a patchwork of retailer-specific connections.
For shoppers, the immediate step is to check an agent’s purchase permissions before enabling it. Spending caps, approval for each order and a clear understanding of the retailer’s return policy can limit unwanted surprises. Agents may make product research easier, but their effect on prices, convenience and customer control will depend on which merchants they can access and how purchases are handled.
Key Questions
What did Walmart announce?
Walmart announced Agent Profiles and an Agent Gateway on October 21, 2026. The gateway is intended to let approved AI agents search inventory and purchase through a structured connection, while profiles can give customers controls such as spending limits.
Why has Amazon blocked Meta’s Muse?
Amazon said Muse had not identified itself as an agent and that its access violated the retailer’s Conditions of Use. Amazon also raised concerns about the agent’s handling of customer credentials. Meta said Muse cannot see users’ passwords or payment methods.
Are AI agents already completing most online purchases?
The supplied evidence does not show that. HUMAN Security reported that, among agent requests it measured in September, 81% went to product and search pages and 2.4% to checkout and payment. That data is one company’s observation, not a market-wide measure.
How could shopping agents affect retailers and publishers?
Agents could change which products customers see and where they check out, shifting influence toward the company operating the agent. Affiliate publishers may also lose credit for sales when an agent uses their product information but completes a purchase without a tracked link. The size of that effect is not yet known.
What should shoppers check before using an agent?
Check whether purchases require approval, set a spending cap if available, and review what account or payment information the agent can access. Confirm the merchant’s shipping, cancellation and return rules. Prices and terms can change, and automated purchasing still carries financial risk.
Source: ThorstenMeyerAI.com
