Personal AI agents — such as Meta’s Muse, Instinct, ChatGPT’s Dots and others — are increasingly able to act on users’ behalf for tasks like booking flights, reserving restaurant tables and ordering groceries. Many users, however, are encountering blocks when the destination website prevents the agent from completing a transaction.
The most publicized example is Amazon, which began blocking Meta’s Muse from parts of its retail site so the agent could no longer browse or make purchases from the product catalog. The issue is broader than Amazon or Muse: social media posts and user complaints indicate other sites and agents are being blocked or interrupted as well.
Why agents get blocked
Some blocks are intentional — as with Amazon — while others stem from traditional anti‑bot measures designed to prevent spam and malicious automated activity. From the user’s perspective it is often unclear whether a block was deliberate or the result of a failed human verification step, such as a button click that the agent cannot perform, which then causes the agent to be ejected.
TechCrunch and other reports note multiple users saying Muse could not complete purchases on Walmart’s site; Walmart told reporters these incidents were not intentional and emphasized that Walmart is a partner of Muse, a partnership announced at Meta’s Connect developer conference in September. Walmart stated it wants to be present where its customers are, including in AI agent experiences.
A common failure mode is a web page that requires a human verification interaction. If that interaction is interrupted or cannot be performed by the agent, the verification fails and the agent loses access.
Different company policies and responses
Not all companies are willing or ready to enable agents. Delta Airlines told reporters it currently has no partnership or integration allowing third‑party AI agents to shop for or book Delta flights on its digital platforms; any opening to agents would have to prioritize “security and the customer experience,” said Heena Chavda, Delta’s General Manager of Global Communications.
United pointed to its Terms of Use that prohibit using “any robot, spider, other automatic device, or manual process to monitor or copy this website” without prior written permission, but did not directly answer whether it was blocking specific personal AI agents.
Other companies have explicit policies. Yelp said it does not permit non‑human traffic on its platform unless the agent has paid access through Yelp’s data licensing program, meaning agents attempting to get quotes, join waitlists or book tables without a partnership will likely fail. eBay said it does not categorically ban all third‑party shopping agents from completing purchases, but its policies restrict unauthorized agents and actions such as automated scraping and data collection for model training. Adidas, Zillow and Pizza Hut did not respond or declined to comment by publication.
Social posts have named other brands reportedly blocking agents, including Yelp, eBay, Zillow, Pizza Hut, Adidas and various airlines; some users say blocking has increased recently. A few users reported that eBay suspended their accounts due to agentic AI use; another said Google ejected Instinct while it was cleaning their Gmail inbox.
Infrastructure providers and the role of CDNs
Cloudflare and similar content delivery networks influence how sites manage AI traffic. Some attribute disruptions in Muse traffic to a change in crawler defaults on September 15 that allowed site owners to block AI training while permitting other classes of bots; as a result, sites previously configured to block AI bots for security reasons were shifted to block AI agents on pages containing ads.
Cloudflare told reporters it had no specific data to share about the blocking of personal AI agents like Muse and pointed to its public data hub, Cloudflare Radar, which shows an overall increase in bot activity but does not differentiate between legitimate and malicious bots. Cloudflare also has commercial reasons to manage AI traffic: it runs a marketplace where AI bots must pay for data access, and it has begun testing a web browser designed specifically for AI agents.
Industry response: open standard for agent commerce
To address these frictions, Meta and partners including Walmart, Stripe, Sierra, Genesys, Rocket, NiCE and Decagon have begun drafting an open standard that would define how AI agents can communicate with businesses. Meta said the protocol will focus specifically on agent‑to‑agent communication for online commerce and aim to distinguish legitimate user‑acting agents from malicious automation.
Meta argued that turning away a personal agent effectively turns away the customer behind it, and prefers to work with businesses to resolve the underlying concerns. Meta also maintains a list of “connectors” (partners) inside the Muse app that is growing; the company says explicitly partnering with brands helps ensure that access problems can be logged and treated as reportable bugs rather than ambiguous blocks.
Why it matters
Personal AI agents are positioned to remove repetitive or time‑consuming tasks for users. If site security configurations, differing company policies, or CDN defaults prevent agents from completing those tasks, user experience suffers and it becomes difficult to identify whether the failure lies with the retailer, the agent provider or infrastructure. An industry standard could reduce that uncertainty, give businesses more predictable control, and create a more reliable way for agents to operate across sites.
Conclusion
The relationship between personal AI agents and websites is evolving rapidly. Some companies are forming partnerships to ensure compatibility, others restrict non‑human traffic, and infrastructure providers’ settings can unintentionally block legitimate agent activity. The open standard under development aims to separate good actors from bad and to create clearer norms for agent‑enabled commerce, but technical and business details remain a work in progress.
