Muse and other AI agents are hitting a wall: websites that block them

Muse users are discovering that some of the web’s biggest destinations won’t let their agent in. According to original reporting from TechCrunch, Meta’s personal AI agent is being blocked, sometimes deliberately and sometimes accidentally, by sites including Delta and United airlines, Yelp, eBay, and Zillow, a new friction point for the agent-driven shopping Meta is building.

Running errands on the open web is Muse’s core pitch. When a site’s defenses treat the agent like a spam bot, the task fails, and the user can’t tell whether the retailer or the agent is at fault.

Amazon’s block was deliberate. Most of the rest is murkier.

The most consequential block is the intentional one: Amazon began blocking Muse from its retail site, so the agent can no longer browse or purchase from its catalog. Bloomberg first reported the move, which came on the heels of Amazon tightening enforcement of its terms against unauthorized agents. Y Combinator co-founder Paul Graham later called the block “the first opportunity” since Amazon’s founding for a startup to build an Amazon competitor, arguing shoppers will want independent agents buying for them rather than Amazon’s own.

Elsewhere, the picture is less clear-cut. Users posting on social media say Muse couldn’t complete purchases on Walmart’s website. But Walmart told TechCrunch the blocks were not intentional; the company is partnered with Muse through an integration announced at Meta’s Connect developer conference in September, and it says it wants to be where its customers are, including inside AI agent experiences.

The likely culprit, according to the reporting, is the human-verification machinery every shopper knows: the “click to prove you’re human” buttons that protect sites from spam. When Muse navigates a checkout, interrupting that verification check can fail it, and the agent gets booted mid-transaction. It’s a sign that the web’s anti-bot defenses, designed for a different era, weren’t built for agents.

Airlines and Yelp draw their lines

Flight booking is one of the headline use cases for agents, and airlines are among the most resistant. Delta told TechCrunch it is taking steps to protect customers from unauthorized automated activity, and that any opening to AI agents would have to be done with “security and the customer experience in mind.” Delta’s general manager of global communications, Heena Chavda, confirmed the airline has no partnership or integration that lets a third-party AI agent shop for or book flights on its platforms, while adding that the company is evaluating how the technology could change how customers engage with it.

United was less direct. A spokesperson pointed to its terms of use, which bar using any robot, spider, or other automatic device to monitor or copy the site without prior written permission, and did not answer whether it is blocking specific agents such as Muse.

Yelp drew the clearest commercial line: the company told TechCrunch it does not permit non-human traffic on its platform unless the agent has paid for access to Yelp’s content and data through its data licensing program. That means agents trying to pull a quote, join a waitlist, or book a table will likely fail without a formal partnership.

eBay took a middle position, saying its approach is not to prohibit all third-party shopping agents from completing purchases; instead, its policies restrict unauthorized agents and actions, such as automated scraping and model training. The company did not address user claims that some accounts were suspended over agentic AI use. Zillow, responding after publication, said it allows AI agents acting on behalf of consumers, though its anti-scraping measures could occasionally block certain agents. Adidas and Pizza Hut did not respond before publication.

The infrastructure layer is part of the problem

Some of the interference may be happening a layer below the websites themselves. TechCrunch noted suspicion among users that content delivery networks, particularly Cloudflare, are disrupting Muse traffic, attributed to a September 15 change in Cloudflare’s crawler defaults that let site owners block AI training bots, which shifted some previously configured sites into blocking AI agents on ad-supported pages.

Cloudflare said it had no specific data to share about the blocking of personal AI agents. The company has a business interest in the outcome: it runs a marketplace where AI bots pay for the data they access, and it has begun testing a browser built specifically for AI agents.

Meta’s answer: an open standard

Meta’s response to the access problem is the industry effort it announced alongside Sierra and partners including Walmart, Stripe, Genesys, Rocket, NiCE, and Decagon: an open standard for agent-to-agent communication in commerce, meant to separate legitimate bots acting on a user’s behalf from bad ones. All Things Muse covered that announcement when it broke.

“Turning away a personal agent means turning away the customer behind it,” Meta said in its announcement of the standards effort. “There is a path that is good for users and good for businesses. It starts with clear norms, gives businesses more predictable control, and evolves toward a more efficient way for agents to work together.”

The company also keeps a list of connectors, its formal partnerships, inside the Muse app, and that list continues to grow. Meta’s argument, in effect, is that partnering is the reliable way in, and that access problems at partner sites are reportable bugs rather than closed doors.

Muse users are living in the gap between that vision and the current web: an agent that can do the shopping, facing storefronts that can’t always tell it apart from the bots they were built to keep out.