Delta Won't Let AI Book Your Tickets. JPMorgan Won't Let AI Move Your Money. The Web Is Closing.
AI companies have spent the last year promising agents that can do anything a human can do online. The web is now actively disproving that.
In recent weeks, users of Meta’s Muse, xAI’s Grok, OpenAI’s Dots, and Instinct have reported blocks preventing their agents from completing tasks across major websites. The list of confirmed blockers now includes Yelp, Delta Air Lines, JPMorgan Chase, Charles Schwab, and Amazon. Cars.com and Amtrak appear to have quietly added blocks without announcement, and user reports suggest others may be following.
Why They’re Blocking
The reasons vary by category.
Financial institutions are citing security. JPMorgan Chase and Charles Schwab have both confirmed they block AI agents attempting to access accounts or move money. Neither has published a detailed policy. The implicit argument — that authorizing an AI agent to transact on a customer’s behalf creates liability they have not been prepared to absorb — is straightforward.
E-commerce platforms are blocking because the economic model breaks down. Amazon does not earn advertising revenue from sessions where an AI agent bypasses its product listings, sponsored placements, and recommendation flows. When Perplexity’s agent shopped Amazon on behalf of users, Amazon claimed it “repeatedly requested” that Perplexity stop. Perplexity countered that Amazon “does not believe in your right to hire labor, to have an assistant or an employee acting on your behalf.”
Data businesses are blocking for licensing. Yelp told TechCrunch it only permits AI agents that pay for its data licensing program. Access to its review corpus has commercial value, and agents extract that value without the transaction that normally funds it.
Airlines are a more direct case. Delta is blocking AI agents from booking tickets. No published policy; no public comment. The inference is straightforward: Delta controls its own booking flow and prefers customers who interact with it directly.
Cloudflare Changed the Default
On September 15, 2026, Cloudflare expanded its block settings so that when a customer enables the block, it applies to mixed-use crawlers including Googlebot, Bingbot, and Applebot. The practical effect: sites that want to stay indexed by search engines but block AI agents can no longer do so cleanly with a single toggle. The tradeoff is now explicit.
Cloudflare also introduced a preset for ad-supported domains that blocks AI agents specifically on pages carrying ads. An ad-supported site that doesn’t want agents reading its content ad-free has a one-click option.
The Amazon Escalation
The blocking is not new, but the scope has expanded. Amazon blocked Perplexity’s agentic shopping feature and filed suit last year. A court issued an injunction against Perplexity. The appeals court overturned the injunction in August 2026. Amazon has since filed an amended complaint.
While that case continues, Amazon has extended practical blocks to major competitors. Users of Meta’s Muse, OpenAI’s Dots, and xAI’s Grok have all reported being unable to complete shopping tasks on Amazon. xAI CEO Elon Musk responded publicly by suggesting he “should start a competitor to Amazon.”
What Meta Did Instead
Meta chose not to fight. In early October the company published a statement: “Turning away a personal agent means turning away the customer behind it.” Rather than attempting to override Amazon’s block, it announced shopping partnerships with Walmart, Gap, Sephora, and Best Buy, enabling Muse to complete purchases on those platforms where access has been negotiated.
The approach frames agents as a distribution channel for retailers rather than as adversaries to them. Whether it scales to the open web is a different question.
The Core Problem
Chris Canal, CEO of AI safety evaluator EquiStamp, described the structural issue: “One of the claims is that they’re going to permeate every aspect of our lives and be able to do anything that we could do online. If there are companies that are actively working to make that not happen, there are just fewer useful things the agents can do.”
The lab position — that an AI agent is an extension of the user, with the same rights to access a website as the user themselves — has not been tested legally at scale. The Perplexity case is still live. No court has established a general principle. In the meantime, every major platform that blocks agents reduces the surface on which the technology’s core value proposition can be demonstrated.
The agentic era assumed a cooperative web. It is getting something more contested.