The Day the Internet Rejected Its Own Assistants
What began as an isolated technical glitch — your AI agent freezing while trying to book a flight or make a purchase online — has rapidly evolved into a systemic movement. Over recent weeks, websites across commerce, banking, search, and publishing have been systematically blocking autonomous agents from accessing their services. And the most striking reality: this is no longer accidental.
What Changed Since September 15
Everything shifted on September 15, when Cloudflare — which handles traffic for a significant fraction of the internet — implemented a new set of stricter patterns for classifying automated traffic. The immediate result: dozens of websites, without their owners asking, started rejecting AI agents as if they were malicious bots. Walmart, for instance, is a partner of Meta Muse, but tested accidental blocks that even prevented real users from making purchases, according to complaints by Kevin Wang on X.
But Cloudflare's change was only the catalyst. The deliberate decision by companies like Amazon, Charles Schwab, and JPMorgan Chase to block third-party agents is the true paradigm shift. Amazon explicitly stated that Meta Muse did not identify itself as an AI agent while browsing its store and that the company had not authorized this integration. Users trying to shop via Muse received a warning: "unauthorized access violates the Terms of Use."
Yahoo News cited Datanyze data showing 20 of the top 100 US websites blocking AI agents. The proportion is even higher among smaller sites, where default security settings are even more aggressive. Google Search has already responded to queries about the topic with a direct block for search agents.
Banking: The Line That Will Not Be Crossed
The firmest response has come from the financial sector. On October 16, Charles Schwab CEO Rick Wurster told Bloomberg Television that his brokerage is actively blocking unauthorized AI agents attempting to move customer money. "We don't want someone just giving their password to Muse and letting Muse come in and act on that client's behalf. We can't control it, the client can't control it, something could go wrong and who's liable?" Wurster said.
JPMorgan Chase, whose CEO Jamie Dimon warned about "rogue agents" during the earnings call, adopted the same stance. Digital Trends noted this is probably the most prudent decision banks have made regarding AI: allowing a chatbot to manage financial transfers without human oversight introduces questions of liability that no company is prepared to answer.
Schwab isn't shutting the door on AI entirely. Instead, the firm is implementing Model Context Protocol (MCP) servers — servers that allow external AI models to access permitted information, such as account balances and market research, but cannot move money.
Who Is Blocking (And Why)
The blocking map is diverse. Each company has a different strategy:
Amazon: Full and deliberate block. The company asked Meta to remove Amazon from the Muse shopping experience and blocked any unauthorized agent. The reason: account security, fraud prevention, and the absence of an authentication protocol that clearly identifies the agent.
Yelp: Paid-access policy. The company stated it does not allow non-human traffic unless the agent has acquired access through its data licensing program. Yelp has turned blocking into a business model: agent access is a product they sell.
eBay: Partial blocking with account suspension risk. The company restricts unauthorized agents and actions like automated scraping and model training. Two users reported that eBay suspended their accounts for using AI agents.
United Airlines: Flight booking blocks. The company did not respond to TechCrunch about whether it is blocking specific personal AI agents, but users report that bookings made by agents are rejected.
Delta Airlines: Blocking policy with ongoing evaluation. Like United, Delta rejects bookings from personal AI agents.
Walmart: Accidental blocking. Despite its declared partnership with Meta Muse, Walmart's site blocked agents due to default security settings. The company told TechCrunch the blocks were not intentional.
Pizza Hut, Adidas, Zillow: All listed by TechCrunch as having agents blocked, though some have not yet issued formal statements.
The Safety Paradox
The irony is that many of these blocks happen without the site owner taking any deliberate action. Cloudflare applies default security policies that, without any extra configuration, start rejecting traffic that looks like it's from an AI agent. Smaller sites, which lack dedicated security teams, end up blocking everything — including agents that are doing exactly what their users ask, in good faith.
This raises a question about the balance between security and innovation: if the internet closes its doors to AI agents in the name of protection, who pays the price? Users who want to use intelligent assistants for everyday tasks get frustrated. Companies that invest in AI agents see their products become unusable on essential platforms. And the sites that block, often without realizing it, miss the opportunity to become part of the agent ecosystem.
The Personal Agent Protocol
In response to the fragmentation, Meta, Amazon, and other companies announced the "Personal Agent Protocol" — an open standard for agent-to-agent communication, especially focused on online commerce. The idea is straightforward: the agent identifies itself, the site responds whether it accepts agents, and if so, under what terms.
The proposal is inspiring. In practice, there is still no widespread adoption. The protocol was announced on October 6, and most major sites have already taken their position — usually "no."
What This Means for the Future of Personal AI
The crisis of blocked agents reveals a fundamental tension: AI agents promise freedom, but they operate on a web that was built to reject automation. The internet as it exists was designed to separate humans from machines — CAPTCHAs, bot filters, Terms of Service. AI agents are, by definition, machines that act like humans.
The future will likely belong to agents that use official APIs and declared integrations, not those that browse the web as disguised scrapers. But that future is still distant. Meanwhile, the internet is becoming an increasingly hostile place for AI agents — and a lesson in how much the web is still, fundamentally, human.
Sources: TechCrunch, Bloomberg, New York Times
✓ Independent sources cross-checked and verified before publishing
