Autonomous AI agents are supposed to scour the internet on your behalf — booking reservations, cleaning out inboxes, buying shoes. But according to The Verge (October 17), the rest of the internet increasingly won't play nice: in recent weeks, users of services like Meta's Muse, xAI's Grok, OpenAI's Dots and Instinct have reported blocks preventing their AI agents from working across top websites.
The named holdouts are specific. 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're blocking AI agents from accessing customer accounts over security concerns. Some users report sites that previously worked with their agents — including Cars.com and Amtrak — are no longer accessible. Sometimes it isn't even deliberate: agents can clash with anti-bot systems even on sites that never formally decided to ban them.
Meta, which builds agents itself, framed it from the agent side: "Turning away a personal agent means turning away the customer behind it." That cuts to the real fight. AI companies want their agents treated as proxies for the user — the same human, just with the keyboard held by software. Platforms see lost ad revenue, unlicensed use of their data, and security exposure, and want the human back in the funnel. It's the same dispute Amazon picked with Perplexity, now spread across airlines, banks, review sites and travel platforms.
How this resolves will shape what the agentic web actually looks like. If enough sites demand licensing deals (like Yelp) or hard blocks (like the banks), agents become services that work only where commercial agreements exist — useful, but a much smaller promise than "an assistant that can use the whole internet." If agent makers win the framing that their agents are simply the user acting, sites face a choice between accommodating them or blocking real customers. Neither side has blinked yet, and more sites are likely to roll out "prove you're human" gates in the meantime.