AI agents might take a bit longer to take over the internet than initially anticipated.
As of this week, Delta is blocking AI agents from booking flights, JPMorgan Chase and Charles Schwab are blocking agents from touching customer accounts, Yelp has put its data behind a paywall for agents, and Amazon has blocked Grok’s shopping bots. That’s just the start of the list, and it’s a lot longer than most people realize.
The common thread is simple: the industry spent the last year building AI agents that can browse, shop, and transact on a person’s behalf, and the industry is now spending this one figuring out it doesn’t want those agents anywhere near its websites.
Amazon vs Everyone
Amazon has been the most aggressive gatekeeper by far. It sued Perplexity in November 2025, accusing the startup’s Comet browser of disguising its AI agent as a regular Chrome user to buy products on customers’ behalf without Amazon’s permission. A federal judge sided with Amazon and granted an injunction in March, only for a Ninth Circuit appeals panel to overturn it months later, ruling that it was the user, not Perplexity, who was technically accessing Amazon’s systems. The case has been sent back to the district court, so the fight isn’t over.
In September, Amazon blocked Meta’s newly launched Muse agent too, greeting users with a popup that read: “Continued access by an unauthorized AI agent violates Amazon’s Conditions of Use, to which our customers have agreed.” Amazon says it asked Meta to voluntarily exclude Amazon.com from Muse’s shopping flow before pulling the plug, and that the agent didn’t identify itself as AI while browsing.
Most recently, users began reporting that Amazon is blocking Grok Bot, xAI’s always-on personal agent, from placing orders on their behalf. One user said a bot he runs just to handle his Amazon returns had stopped working. Elon Musk’s response was to float the idea of a rival e-commerce platform altogether: “Maybe we should start a competitor to Amazon.” Amazon hasn’t confirmed the Grok-specific block the way it did with Muse, but the pattern fits a company that has now gone after three separate agent products in under two months.
Banks Are Drawing The Same Line
Charles Schwab and JPMorgan Chase, two of America’s largest financial institutions, both announced that they’re blocking unauthorized AI agents from accessing customer accounts, specifically naming Meta’s Muse among the agents they stop.
Schwab CEO Rick Wurster told Bloomberg Television that the company blocks agents trying to move client money “tens of times a week,” and that he gets a report on it. Instead of letting outside agents log in with a customer’s credentials, Schwab is building its own controlled MCP servers and pushing users toward its in-house assistant, Charley.
JPMorgan CEO Jamie Dimon was blunter about it on the bank’s Q3 earnings call: “You don’t want rogue agents, and it could be thousands of them.” A JPMorgan spokeswoman confirmed the bank already blocks external agents from operating on behalf of retail customers, though agents can still see read-only data like account balances through providers like Plaid. Dimon also raised the question of who’s liable if a customer hands an agent their banking credentials and something goes wrong.
The banks aren’t being cagey about their motive either. With idle cash sitting in low-yield accounts, an agent that can shop around for a better interest rate and move money automatically is a direct threat to a business line both companies care about.
Yelp’s Approach: Pay To Play
Yelp’s position is a little different from the others. Its robots.txt file disallows crawling by default, and its terms of service explicitly prohibit scraping, so agents trying to read reviews or business listings without permission get blocked the same way any other scraper does. But Yelp isn’t refusing agents outright, it’s charging for the privilege. The company has struck data licensing deals with OpenAI, Perplexity, and other AI platforms, reportedly pulling in tens of millions of dollars a year from the arrangement, on top of existing partnerships that already feed Yelp data into Amazon Alexa, Apple Maps, and Microsoft Bing.
So where Amazon and the banks are mostly playing defense, Yelp has turned agent access into a revenue line. Early reports from TechCrunch also named eBay, Zillow, Pizza Hut, and Adidas among brands that users have complained about being blocked by, suggesting Yelp’s posture, charge for access rather than block it outright, may become the more common playbook as more sites figure out what their data is worth.
What About Delta
Delta’s case is the murkiest of the bunch. Delta hasn’t issued an outright block the way Amazon or the banks have. When TechCrunch asked directly whether Delta blocks AI agents, the airline’s General Manager of Global Communications said Delta “currently has no partnerships or integrations that allow third-party AI agents to purchase or book Delta flights on behalf of customers,” while adding that it’s “continuing to actively evaluate” the technology. In practice, that’s meant users reporting their booking agents getting rejected, which functions like a block even without a formal policy.
Delta is also hedging its bets by testing its own browser built specifically for AI agents, a sign that airlines, like the banks, would rather build the agent-facing product themselves than let a third party sit between them and the customer.
Why This Is Happening Now
None of this is really about security, even though that’s the language every company reaches for. It’s about who owns the customer relationship, and the ad and data revenue that comes with it. Amazon makes tens of billions of dollars a year from sponsored placements and recommendations; an agent that skips straight to checkout bypasses all of it. Banks earn on idle cash sitting in low-yield accounts; an agent built to chase yield threatens that. Every platform an agent is “disrupting” has an incentive to keep that agent out and replace it with one of its own.
Cloudflare, which sits in front of roughly a fifth of the internet, has been building the infrastructure for exactly this standoff, rolling out tools that let any website block AI crawlers by default and charge the ones it does let in through a “pay per crawl” system. What started as a defense against AI training scrapers is increasingly being repurposed for live shopping and booking agents too.
The legal question underneath all of it, whether a company can stop an AI agent from doing something a human user explicitly asked it to do, is still unsettled. The Ninth Circuit’s Perplexity ruling leaned toward the user’s side; Amazon’s unresolved lawsuit and its newer blocks on Muse and Grok Bot suggest it isn’t backing down. Expect a lot more of this before anyone agrees on the rules.