Major websites, airlines and banks are closing their doors to AI agents, just as those agents are learning to browse, book and buy on behalf of users. In recent weeks, users of Meta’s Muse, xAI’s Grok and OpenAI’s Dots and Instinct have reported being blocked on major sites, according to a report by The Verge published on October 17, 2026.
The blocks are not accidental glitches. Companies are stating the policy in plain terms, arguing that third-party agents using customer credentials or browser automation create security, liability and commercial risks they are not prepared to accept without a formal deal.
What Yelp, Delta And The Banks Are Saying
Yelp told TechCrunch it allows AI agents only if they have paid for access through its data licensing program. That turns agent access into a licensed product, rather than free browsing of public listings built for human visitors.
Delta Air Lines is preventing agents from booking tickets on users’ behalf. Delta says it does not currently have a partnership or integration letting a third-party agent shop for or book its flights, is continuing to evaluate the area, and is taking steps to protect customers from unauthorized automated activity.
At JPMorgan Chase, chief executive Jamie Dimon said on the October 13 earnings call that the bank will not allow just any outside agent inside, because it is too risky for clients and the bank. A spokesperson confirmed the bank blocks outside AI agents from performing tasks for customers.
Charles Schwab has taken the same line. Chief executive Rick Wurster told investors on October 15 that an agent logging in on behalf of a client is not safe and will not be allowed, and Schwab blocks agents using client login credentials or browser automation.
How The Blocking Works In Practice
Most consumer agents work by controlling a browser, reusing a signed-in session, or automating clicks and form fills in the way a human would. That makes them hard to distinguish from other automation, and easy to stop with controls sites already use against bots.
Even without an explicit ban, login screens, CAPTCHAs and device checks trip agents up and break multi-step tasks. Cloudflare, which sits in front of a large share of the web, began offering one-click blocking of AI crawlers in 2025, giving site owners a simple switch for automated access.
Some users say Cars.com and Amtrak, which previously worked with their agents, are no longer accessible to them. Those cases are reported by users and remain unconfirmed as formal bans, unlike the on-record positions from Yelp, JPMorgan Chase and Schwab.
The Partnership Route Around The Wall
There is a clear counter-example to the blanket block. Meta has partnerships with Walmart, Gap, Sephora and Best Buy that let Muse shop on their sites, showing retailers will welcome agents when access is negotiated, attributed and commercially agreed.
Shopify points the same way. The company has opened its storefront network to AI agents, so merchants can be reached through agents under platform rules instead of through unauthorised scraping or session reuse.
The divide is therefore not between sites that accept agents and sites that reject technology. It is between licensed application programming interfaces and partnerships, with identity, consent and payment terms defined, and unauthorised browser automation that sites treat as a security risk. Readers new to the distinction can start with what AI agents are and how AI agents differ from chatbots.
Why It Matters
AI agent web access is the central issue here, because an agent is only useful if websites, banks and booking systems let it act. Its key attributes are authentication, whose credentials it uses, permission, whether the site has consented, licensing, whether access is paid or partnered, and detection, whether anti-bot systems can identify and stop it.
Its neighbours explain the stakes. Agent platforms, covered in our guide to agent platforms to know in 2026, need broad reach to complete tasks, while banks, airlines and marketplaces carry fraud, ticketing and data duties that make open access costly if an agent acts wrongly.
If partnership becomes the default gate, agent capability will depend less on model quality and more on commercial deals. Users may find their chosen agent works at one retailer or airline and fails at another, not for technical reasons but because no agreement exists.
Background
Websites have fought automated access for decades, from early search crawlers and price scrapers to modern bot-management systems built to stop credential stuffing and ticket scalping. AI agents inherited that contested ground when they moved from answering questions to using the web directly, and the current blocks are the first broad, on-record response from banks, travel firms and platforms to that shift.
Source: The Verge report via The Tech Street Now, RR Tech Pulse
0 Comments