Amazon Draws a Line on Meta’s Shopping Bot as Retailers Push Back on AI Agents

Amazon has blocked Meta’s Muse shopping agent from browsing its site, a sharp sign that retailers are no longer treating AI shopping bots as a novelty. The move, reported by The Verge, underscores a growing struggle over who gets to control online shopping: platforms and merchants, or the software acting on behalf of customers.

Amazon’s block marks a clearer retail red line

According to The Verge, Amazon prevented Meta’s Muse from shopping on its site, even as the social media company has positioned the tool as a way for users to browse with permission and step in whenever they want. The same report says Amazon has already sued Perplexity over shopping activity on Amazon after asking the company to stop its agent.

That combination of technical blocking and legal pressure suggests Amazon is not merely trying to manage traffic. It is signaling that autonomous or semi-autonomous buying tools will be allowed only on its terms, if at all.

Amazon did not frame the issue as an anti-AI stance. Spokesperson Maxine Tagay said the company had “requested that Meta remove Amazon” and added that “third-party applications that offer to make purchases on behalf of customers from other businesses should operate openly.”

Why it matters

AI shopping agents promise convenience: they can search, compare, and in some cases try to complete purchases with less work from the customer. But the retailer response now taking shape shows those benefits may be constrained by platform rules, anti-bot defenses, and business incentives that favor direct customer relationships.

For shoppers, that could mean a patchwork experience. An agent may work on one site, stall on another, and fail entirely on a third. For retailers, the stakes include pricing control, checkout integrity, inventory fairness, and the risk that automated tools will create traffic patterns or transaction behavior the seller cannot easily verify.

For the broader market, the dispute is an early test of whether “agentic commerce” becomes a normal layer of online retail or a fenced-off feature limited to sites willing to cooperate. The answer could shape how consumers discover products, how sellers pay for access, and how much power the biggest platforms retain over the path to purchase.

How retailers are responding

The reaction is not uniform. The Verge reported that Target allows some AI agents to browse parts of its site, but blocks unauthorized access and transactions. Target also said agent success depends on the task path and the safeguards built into the site. The company said it has partnerships with three major AI firms for shopping tools: OpenAI, Google, and Microsoft.

That stance points to a selective model rather than a blanket welcome. Retailers may be willing to host outside agents when those tools respect site rules, but they appear far less willing to let them roam freely through checkout or transaction flows.

JPMorgan Chase is even more restrictive. The Verge reported the bank will not allow outside agents to access accounts, citing risk. While banking is not retail, the logic is similar: once software can act with authority, the company behind the system wants to know exactly what it is doing and whether it should be allowed at all.

Even where there is no explicit ban, basic bot defenses can still break the experience. The Verge said general protections against automated traffic may interfere with agent shopping even when a site has not formally prohibited it. That means the technical layers built to fight scraping and abuse can also slow or block legitimate consumer tools.

What happened when Muse went shopping

Meta’s Muse was tested across several retail environments, and the results were uneven. The Verge reported that Muse was intermittently blocked by Target during shopping tests. On Amazon, the agent could suggest alternatives when the site was unavailable, but it could not confirm Amazon pricing. That is a crucial limitation for any shopping assistant that claims to help users compare offers or make decisions based on current costs.

At The RealReal, the experience was more accommodating. The company told The Verge it is open to third-party agents while still guarding against disruptive automation. Chief technology officer Luke Friang said the company “remain open to third-party AI agents.”

The RealReal’s setup is notable because its cart reservation system can help agents hold items temporarily. That matters on a secondhand luxury marketplace, where inventory is limited and speed can decide who gets an item. In that environment, a bot that can add to cart and preserve access for a short window may be useful to a legitimate shopper — but it also creates a new question about fairness when many shoppers, human and machine, are chasing the same item.

Meta’s pitch for Muse

Meta says Muse can browse with permission and that users remain in control. Tracy Clayton, a Meta spokesperson, said “users can take over at any time.” The company also says retailer response to Muse has been overwhelmingly positive.

That upbeat framing is in tension with the retailer pushback documented by The Verge. If merchants are increasingly blocking or limiting the agent, then the user experience Meta is promoting depends heavily on where the tool is allowed to operate and how robustly a site defends itself against automation.

In practical terms, an AI shopping agent only works as well as the supply of sites that will tolerate it. Once a major retailer refuses access, the promise of seamless cross-site shopping becomes much harder to deliver.

Background: how AI shopping agents emerged

AI shopping agents are software tools designed to automate parts of online buying. They can browse product listings, compare choices, and sometimes move through checkout steps on a user’s behalf. The idea fits into a broader wave of AI products that seek not just to answer questions, but to take actions.

That shift creates a direct collision with how retail websites are built. Most platforms were designed around human browsing patterns, human login behavior, and human purchasing pace. Retailers have long used bot protections and account safeguards to limit automated access, both to prevent abuse and to protect the integrity of shopping and inventory systems.

Amazon’s move is especially significant because the company also runs its own agentic shopping features on-platform, according to the background in The Verge’s reporting. That means the fight is not about whether AI can assist shopping at all. It is about which AI assistants get to participate, under whose rules, and with what level of visibility.

Retail and platform strategies at a glance

Different companies are drawing different lines. The table below summarizes the positions reported by The Verge and the implications of those choices.

By the numbers

The numbers in this story are modest, but they reveal how specific the practical issues can be when shopping agents meet real retail systems.

What the friction tells us about the future of checkout

The tension around Muse is not just technical. It is commercial. Retailers worry that autonomous agents may scrape information, trigger unwanted traffic, misuse carts, or distort how purchases are initiated. Consumers, meanwhile, are being told that AI can reduce the burden of online shopping. Both things can be true at the same time.

The result is likely to be fragmentation. Some stores may allow agents in controlled ways. Others may permit browsing but not transactions. Some may block them outright. And a few may build their own agent tools, keeping the experience inside their own platforms.

That arrangement could blunt the convenience story for users. Instead of one universal shopping assistant, customers may end up with a set of site-by-site permissions, each with different rules and reliability. The more restrictive the major platforms become, the more likely agent shopping becomes another feature trapped inside walled gardens.

There is also a competitive dimension. If a retailer allows outside agents but a rival blocks them, the first may gain traffic from customers who value automation, while the second preserves tighter control over its own ecosystem. That creates a new battleground over access that goes beyond classic search rankings or app-store placement.

What happens next

The next phase will hinge on which companies decide to set formal rules for third-party agents and which ones continue to rely on technical barriers. The Verge’s reporting leaves several questions open.

• Will more major retailers formally allow or ban outside shopping agents?
• Will lawsuits or cease-and-desist demands become a standard response to agentic shopping?
• Can platforms balance customer convenience with concerns about scraping, bot abuse, and purchase integrity?

Those questions matter because the present moment looks less like a smooth rollout than a negotiation. Retailers are testing the limits of AI shopping tools one site at a time, and the rules are being written in public.

For now, the clearest takeaway is that AI shopping agents are no longer being judged only on how well they work. They are being judged on whether the businesses behind the websites want them there in the first place.

Bottom line

Amazon’s decision to block Meta’s Muse from shopping on its site is more than a single platform dispute. It is a sign that retailers are beginning to define the boundaries of AI-assisted commerce, and those boundaries may determine how much of the online shopping experience is eventually handed over to machines.