United Airlines wrote AI agents into its contract of carriage as a prohibited practice: what the text says and what retailers should take from it
Supervision is no defence, searching is still allowed and United can cancel without notice; the first airline ban on agents is a template other merchants will copy or reject.

United Airlines revised its contract of carriage on November 11 to prohibit AI agents from making, modifying or servicing tickets or buying any product, and to let it cancel reservations made that way without notice. It is the first major US carrier to put the rule in the contract passengers accept when they buy a ticket; Delta was reported in October to be blocking similar tools technically. The Wall Street Journal first reported the change.
What the text says
A new definition in Rule 1 describes an AI agent as software or services, including AI assistants, agents or AI-enabled browsers, that take autonomous or semi-autonomous action to discover, search, book, purchase, ticket, service or manage United products on behalf of a passenger or anyone else, whether or not a person supervises or approves what the software does. The examples cover assistants built into social, messaging and search platforms, AI browsers and standalone agents, and third-party assistants reaching United through a travel agency's tools. Rule 6.J.6 adds the use of such agents to the list of prohibited ticketing practices, alongside hidden-city and throwaway ticketing. Rule 5.H adds reservations made, modified or serviced by agents to the bookings United may cancel without notifying anyone. Existing remedies for prohibited practices include voiding tickets, charging the full fare after travel, deleting miles, revoking status and terminating MileagePlus membership. Bookings made before November 11 are not affected, assistive technology for passengers with disabilities is exempt, and travel agencies may use AI tools only with United's permission.
What is still allowed
The definition covers discovering and searching, but the prohibition covers only making, modifying or servicing a ticket and purchasing. Using an assistant to find and compare fares and then buying on united.com yourself is permitted, which is also the model Expedia's ChatGPT integration already follows: show the options, send the customer to checkout. United has not defined 'service a ticket', has not said how it will detect agents, and says agent bookings are rare today and that it works with no third-party agents.
United's reasons, and the other one
United cites three: passengers sharing logins with third-party tools expose personal data; agents can make unintended purchases, account changes or spend miles; and federal dangerous-goods rules require a passenger to personally acknowledge warnings at purchase and check-in, which an automated check-in would bypass. Chief executive Scott Kirby reportedly gave a fourth at an event in Chicago: he did not want to create another middleman. Analyst Henry Harteveldt told CBS that searches cost airlines money and agents generate searches that do not convert, which is the same economics that pushed Amazon to cut off Meta's Muse agent in September and win an injunction against Perplexity's Comet in March. Walmart went the other way in October, publishing terms and spending limits under which outside agents may transact.
The counter-argument
A contractual ban is not a technical one. A booking record may show Expedia as the seller and nothing about the assistant that drove it, and United's own definition catches a customer who approves every step, which is the behaviour the HUMAN Security survey in June found 54% of Americans comfortable with for planning and 12% for autonomous booking. Punishing the customer's loyalty account for a tool choice is the part most likely to be tested, in court or by the Department of Transportation, which has already asked United, Delta and JetBlue to explain how they set fares. And the last industry attempt to police booking channels, American's move to deny miles on some agency bookings, was walked back.
What to do
If you build agents: implement a merchant policy check, stop at fare display on united.com, and log the hand-off so you can show a customer did the buying. If you run a transactional site: write down which of the three postures you are taking, publish it in your terms, and instrument agent traffic now; Arkose reported in October that 97.7% of legitimate agent sessions carried no verified identity, so detection is a product decision, not a configuration. If you are a traveller: use AI to search, buy yourself, and do not hand your MileagePlus login to a browser extension. Watch for Visa's Trusted Agent Protocol and similar schemes, because a verified, paying agent is the version of this that airlines will eventually welcome.
More in Enterprise & business
All →
OpenAI's $10 billion quarter came with a $10.9 billion loss: what the leaked numbers say next to Anthropic's S-1
Revenue up 50% in a quarter, enterprise now bigger than consumer, and a loss of $1.09 per dollar of sales; the figures explain the Cursor cutoff, the …

Claude Sheets, Whiteboards and Meetings: Anthropic now sells an office suite, and here is what IT buyers should ask before the renewal
The day before its roadshow, Anthropic turned Claude from a chat window into a workspace with spreadsheets, shared boards, embedded partner apps and a…

Google's four TPUs in one year: what 8tx and 8i Max mean for anyone negotiating with Nvidia
Confidential compute for fine-tuning and a doubled-up inference chip extend the eighth-generation line, and the software around it is now the real com…
Get the brief in your inbox
One email every morning: the stories that mattered in AI, each with a one-line "why it matters" for your role. Free, no ads, unsubscribe any time.
Published daily by AI Briefing. We never share your address.