Remember when every keynote promised that your AI assistant would soon book your flights, pick your seat and argue with the airline for you? Well, United Airlines just answered that pitch with a polite but firm “no, thanks”. As of this week, United bans AI agents from booking flights on behalf of travellers, and it’s rewriting its booking and ticketing policies to make it official.
The airline’s reasoning is a mix of security, lawyering and, let’s be honest, money. In its statement, United says that “sharing account credentials with personal, third-party, AI assistants can expose sensitive information”, that an agent could make purchases you didn’t intend, mess with your account or burn your MileagePlus miles, and that letting a bot check you in could even break federal rules, since you’re supposed to personally confirm that your bag doesn’t contain anything explosive. A chatbot swearing there’s no dynamite in your carry-on apparently doesn’t count.
Why United bans AI agents (the real reason)
Those arguments aren’t nonsense. An agent that holds your login and your credit card and clicks “confirm” on your behalf is a juicy target, and anyone who has watched one of these things hallucinate a date knows a wrong booking is one bad token away. But the analyst quoted by CBS News, Henry Harteveldt of Atmosphere Research Group, puts his finger on the less noble part: “United is saying they don’t want to add yet another, and potentially even more powerful, intermediary.” And then the kicker: “Flight searches cost airlines money.”
That’s the thing nobody says in the keynotes. Every time an agent hammers a booking engine to compare 40 itineraries for you, the airline pays for the compute and gets nothing back. Airlines spent two decades fighting online travel agencies for control of the customer relationship. They are not about to hand that relationship to OpenAI or Meta for free.
Funny detail: United admits it hasn’t actually seen many people trying to book this way yet. DoorDash said the same thing recently. So this is a pre-emptive strike, a fence built before the cows show up. Meanwhile Southwest went the other direction and made its inventory visible to ChatGPT users. Two airlines, two bets, and I suspect we’ll find out within a year which one looks smart.
What it means for your AI assistant
Practically, not much changes for you today. United fares still show up on Kayak, Hopper and Google Flights, and your agent can still look at United flights; it just can’t push the buy button. As Harteveldt notes, consumers “shouldn’t find it any more difficult to find good fares”, but “by blocking the AI bots, United won’t make it any easier to find its fares, either.”
The bigger picture is more interesting. We’re watching the first real pushback against the “agentic web” from companies that actually sell things. Publishers already told the bots to get lost with a new robots.txt signal; now merchants are drawing their own lines, and they have a much better weapon than a text file: they can simply refuse the transaction. Expect more of this. Banks, ticket sellers and anyone with a loyalty program have exactly the same incentives as United.
My take? I’d never hand an AI agent my airline password anyway, and if you’ve done it, go change that password tonight. Use the bots to research, then do the last click yourself. It’s slower, sure. It’s also the one part of the trip you really don’t want automated by something that can’t be held responsible when it books you to Portland, Maine instead of Portland, Oregon.
Your assistant can plan the trip. It still can’t fly without you, and now it can’t book without you either.
Discover more from TechPassion
Subscribe to get the latest posts sent to your email.
