AI agents that browse the web and act for their users are running into a wall of their own: popular services are blocking them. The Verge reported on October 16 that users of agents from Meta, xAI, OpenAI and others have seen tasks fail across top websites, and that several large companies have confirmed they restrict agent access deliberately.

Who is blocking what

According to The Verge, citing statements the companies gave to it and to TechCrunch:

  • Yelp told TechCrunch it only allows AI agents that pay for its data licensing program.
  • Delta Air Lines is preventing agents from booking tickets on behalf of users.
  • JPMorgan Chase and Charles Schwab confirmed they block AI agents from accessing accounts because of security concerns.

Some users also say that sites which previously worked with their agent, including Cars.com and Amtrak, are no longer accessible, which suggests other companies may be blocking agents without announcing it. Those reports are anecdotal, and the companies named have not confirmed any change. The article lists agent products from Meta (Muse), xAI (Grok), OpenAI (Dots) and another company called Instinct among those affected.

Three different motives

The reasons behind the blocks vary, and they matter for how the situation might resolve.

Revenue and data value. Yelp’s stance is a pay-to-play model: if an agent uses its reviews, it should be part of a licensing deal. This mirrors how publishers have approached AI crawlers.

Security and liability. Banks and brokers worry about unauthorized access to accounts and about who is responsible when an agent makes a mistake or is tricked. Agents that read web pages can be manipulated by hidden instructions, a risk explained in our piece on prompt injection and LLM security.

Business model. Sites that earn from human attention, such as through advertising, lose out when an agent skips the page views. The Verge recalls that Amazon banned Perplexity’s agentic shopping feature last year and sued the AI startup, arguing that third parties acting for customers “should operate openly and respect service provider decisions.” Perplexity replied that Amazon does not want agents because it would see fewer ads, and compared its agent to an assistant acting on a user’s behalf. A judge ordered Perplexity to stop its agents from shopping on Amazon, but an appeals court overturned that in August, and Amazon has since filed an amended complaint. The case is still unresolved.

The technical wall

Even sites that do not ban agents outright can break them with anti-bot defenses such as challenge pages and behavioral detection, which are designed to stop automated traffic. An agent that acts like a person in a browser can be hard to tell apart from a scraper, so legitimate user-directed activity gets caught in the same net. That pushes the industry toward explicit, sanctioned channels: APIs, partner programs and standards for authorizing an agent, such as the Model Context Protocol described in our MCP explainer.

Why it matters

The promise of AI agents is that they can handle errands across the whole web. If major services refuse them, agents become useful mainly where companies opt in, which favors big platforms that can strike deals and puts independent agent developers at a disadvantage. For consumers, it raises a basic question about whether a person has a right to delegate tasks to software on sites they are allowed to use. There is no settled answer in law or policy; the Amazon–Perplexity litigation may offer the first signals.

The practical effect today is more fragmentation: some agents will work on partnered sites, others will not, and users may find failures that offer no clear explanation. Our overview of what AI agents are gives the background for readers new to the topic.

What to watch next

  • The next court ruling in Amazon’s case against Perplexity.
  • Whether Yelp-style licensing programs spread to other review and listing sites.
  • New authorization standards that let sites approve specific agents safely.
  • Whether banks offer official, secure agent access for account tasks.
  • Disclosure by agent makers about which sites block them.

Key takeaways

  • Yelp, Delta, JPMorgan Chase and Charles Schwab have confirmed they limit or block AI agents.
  • Motives range from licensing revenue and security to protecting ad-driven traffic.
  • Reports of silent blocks on other sites are anecdotal and unconfirmed.
  • Amazon’s lawsuit against Perplexity over agentic shopping remains unresolved.
  • Sanctioned access, such as APIs and partner programs, is the likely long-term path.

Frequently asked questions

Why would a website block an AI agent that is helping a customer?

Because the site may lose ad views, data licensing revenue or control over security and liability. Each company’s reasons differ.

Sites generally set their own terms of access, but the law on user-directed agents is unsettled and under litigation.

Will agents stop working on these sites for good?

Not necessarily. Many sites may allow agents through approved integrations or paid programs rather than open browsing.