United Bans AI Agents From Booking Flights, Nov 11

United Airlines doesn’t want your AI agent anywhere near its checkout page anymore. On November 11, the same month Meta’s Muse kept getting blocked from retail sites, United added a new contract rule, 6.J.6, barring AI agents from booking or buying anything through its site or app. A companion rule, 5.H, lets United cancel a booking it decides an agent made.

The timing isn’t random. United’s own newsroom post says sharing account credentials with “personal, third-party, AI assistants” can expose sensitive information and lead to “unintended purchases, account changes, or use of MileagePlus miles.” It also raises a regulatory point nobody else in this fight has used yet: flight check-in requires a human to confirm their bags don’t contain banned items like explosives, which an automated agent clicking through on someone’s behalf can’t meaningfully do under FAA and DOT rules.

What the rule actually bans

United’s definition is broad on purpose. It covers AI assistants built into messaging or search apps, AI-enabled browsers, and standalone agents, whether or not a person is watching or approving each step. It also reaches indirect access: travel agencies and booking platforms can’t let a third-party AI agent search, book or manage a United reservation for their customers either, unless United has separately agreed to it.

What it doesn’t ban is research. United’s Digital Trust Center page gives the example of asking an AI tool “What are the best things to do in New York in May?” as fine, while asking it to pull up a confirmation number and upgrade a seat is not. That distinction matters, because it’s the same line Expedia’s ChatGPT integration already sits on: the chatbot can show flights and prices, but the purchase itself happens on Expedia’s own site, not inside the chat. United says it isn’t blocking that kind of comparison shopping, only agents that try to transact directly.

Bookings made by an unauthorized agent before November 11 won’t be cancelled retroactively, United says, though it’s telling customers to double check anything an agent booked for them. Anyone who shared MileagePlus login details with a third-party agent is told to change the password now.

The part that’s easy to misread

This isn’t United deciding AI has no place in travel. It’s United deciding that an agent holding your password and clicking “buy” is a liability it won’t carry, at least not yet. The newsroom post explicitly frames the move as following banking and retail, saying “corporate leaders in the banking and retail industries have taken similar actions regarding AI agents and we encourage others to consider measures that preserve the data and security of their customers.”

That line is the one worth sitting with, because retail didn’t actually move as one block. In September, Amazon blocked Meta’s Muse agent from its site, saying Muse didn’t identify itself as a bot and appeared to capture login credentials. Eight days later Shopify did the opposite: it shipped WebMCP checkout tools that let a signed, identified agent complete a real purchase on any participating Shopify store, with Shop Pay handling the payment and a human still confirming at 3D Secure. (I wrote about that split in September; see the United/Shopify comparison table below.)

United’s new rule reads much closer to Amazon’s position than Shopify’s, which opened its own checkout to signed agents in late September. The difference is identity and scope, not the idea of an agent buying something at all. Shopify’s tools require a registered, cryptographically signed agent and never hand over raw card details. United hasn’t built an equivalent signed-agent lane, it’s just closed the door until one exists, if it ever does. The airline says flatly on its FAQ page that it doesn’t currently partner with any third-party AI agent.

	What’s allowed	What’s blocked
United (Nov 11, 2026)	Asking an AI tool general travel questions, Expedia-style chat that redirects you to buy yourself	An agent logging into MileagePlus, booking, upgrading, checking in, or buying on your behalf, even with your permission
Shopify (Sept 28, 2026)	A signed, registered agent completing checkout with a saved Shop Pay card	An agent entering new card details or bypassing a 3D Secure step
Amazon (Sept 2026)	Agents that identify themselves and get authorization	Unauthorized agents that don’t disclose they’re bots

My read: the FAA/DOT bag-check-in point is the strongest part of United’s case, because it’s not really about trust in AI, it’s a rule a human has to clear no matter how good the agent gets. The privacy argument is weaker as a permanent position. Shopify showed a signed-agent, limited-payment model can exist within weeks, and airlines run loyalty programs with far more valuable fraud surface (free flights, upgrades) than a retail cart. I’d expect United’s “we don’t partner with anyone yet” line to have an expiration date once an agent vendor builds the identity and liability layer airlines are asking for. The travel-agency clause is the one to watch next: if a major booking platform like Expedia or a flight-search agent negotiates an approved integration, that’s the real signal this is a pause, not a permanent wall.

Note on sources: I read United's Digital Trust Center page and its newsroom release in full. PhocusWire, InsideFlyer and the Verge's reporting were used to confirm the contract rule numbers and effective date; I did not independently verify United's internal fraud or booking data, which the airline hasn't published.

United banned AI agents from booking flights. Two months ago, Shopify invited them in.

What the rule actually bans

The part that’s easy to misread

Sources