Consumer AI agents promised a simple pitch: book the flight, make the reservation, order the groceries — automatically, on your behalf. But early adopters are discovering that the hardest part of agentic AI is not the AI. It's getting websites to let the agents in.
According to a report from TechCrunch, Amazon has begun blocking Meta's Muse AI agent from its retail site, preventing the agent from browsing or making purchases from Amazon's product catalog. And the problem extends well beyond Amazon: users across social media report their agents being stymied by retailers, airlines, and review sites — sometimes intentionally, sometimes as collateral damage from anti-bot defenses. For more context on this story, see our ongoing latest AI developments.
The Growing List of Sites Blocking AI Agents
The friction points reported by users and confirmed by TechCrunch's reporting span much of the consumer web:
- Amazon began blocking Meta's Muse agent from browsing and purchasing on its retail site.
- Walmart customers have complained that Muse could not complete purchases, though Walmart says the blocks were not intentional. The issue often traces to human-verification buttons: if a user doesn't complete the check in time, the session fails and the agent gets booted.
- Delta told TechCrunch it takes "steps to protect customers from unauthorized automated activity," and that it currently has no partnership enabling third-party AI agents to shop for or book flights on customers' behalf.
- United pointed to its Terms of Use, which bar users from employing "any robot, spider, other automatic device, or manual process" on the site without prior written permission.
- Yelp said it does not permit non-human traffic unless an agent has paid for access through its data licensing program — meaning agents trying to grab a quote, join a waitlist, or book a table will likely fail without a formal partnership.
- eBay said it does not prohibit all third-party shopping agents, but restricts unauthorized agents and actions like automated scraping and model training.
User complaints compiled by TechCrunch also reference eBay, Zillow, Pizza Hut, Adidas, and various airlines blocking agents, with some users reporting that eBay suspended their accounts over agentic AI use. One user said Google kicked the Instinct agent out mid-task while it was cleaning up their Gmail inbox. Some users suspect Cloudflare and similar content delivery networks — which sell protection against automated traffic — are disrupting Muse traffic as well.
When Blocking Is Intentional — and When It Isn't
The Walmart case illustrates how confusing the landscape has become. Walmart is, in fact, a Muse partner: the collaboration was announced at Meta's Connect developer conference in September, and the retailer told TechCrunch it wants to be available through AI agent experiences. Yet shoppers still hit verification screens that agents cannot pass, leaving humans frustrated with both the retailer and the agent.
For a spokesperson's statement and Delta's measured response alike, the subtext is the same: companies want to control how agents interact with their platforms, and few have settled on terms they're willing to publish. In the absence of formal agreements, an agent that looks like a bot gets treated like one.
Meta, Walmart, Stripe, and Others Draft an Open Standard
The industry's response is taking shape. According to TechCrunch, partners including Meta, Walmart, Stripe, Sierra, Genesys, Rocket, NiCE, and Decagon have begun work on an open standard dictating how AI agents can communicate with businesses — helping distinguish trusted agents acting on a user's behalf from malicious bots.
Meta said in a blog post that the protocol will focus specifically on agent-to-business communication for online commerce. If it succeeds, the standard would function something like a passport system for agents: verifiable credentials that let a website know an agent is who it claims to be, acting for a real customer, within agreed limits.
That would address the core problem from both sides. Websites gain a way to admit the traffic they want without weakening anti-fraud defenses, and agent makers gain a reliable path to complete purchases that doesn't hinge on outsmarting CAPTCHAs.
Why the Stakes Are Higher Than a Blocked Purchase
The agents now hitting walls — Meta's Muse, Instinct, ChatGPT's Dots, and others — represent a major bet by the industry that commerce will shift from humans clicking through sites to agents executing tasks. If the open web's default answer to agents is "no," that vision stalls, and the value of agentic features shrinks to the handful of sites that opt in.
There is also a consumer-confusion cost. When an agent fails mid-purchase, the user often cannot tell whether the retailer blocked the agent on purpose or the site's bot defenses misfired — a frustration that lands on both brands at once.
The standoff between retailers and agents is, in effect, a renegotiation of the web's terms of service for the agentic era. Amazon's block on Muse shows platforms still hold the leverage. The open standard now in the works is the industry's attempt to replace ad-hoc blocking with rules both sides can build on. Whether it arrives fast enough — and wins adoption beyond its founding partners — will shape how quickly AI agents move from demos to daily life.
---
Stay Ahead of AIGet the latest AI news, analysis, and breakthroughs — all in one place.
Read more AI news →
