United Airlines has drawn a line around the final, most consequential step of online travel: booking the ticket.

On Wednesday, the carrier said third-party AI agents acting as personal assistants would be barred from accessing customer accounts and completing flight reservations. The move arrives as Meta, Google, OpenAI and Anthropic push assistants beyond answering questions and toward carrying out everyday tasks such as shopping, restaurant reservations and travel planning.¹

United is not shutting AI out entirely. Agents can still surface the airline’s flights, but they cannot finish a purchase for a customer. The company’s concern is that an autonomous tool could expose private account information, make an unintended purchase or bypass federally required disclosures and confirmations in the booking process.²

That last issue is especially awkward for aviation. Airlines must collect and present specific safety information, including passenger attestations related to prohibited items in baggage. An AI agent cannot simply stand in for a traveler on such declarations, according to United’s rationale.²

Chief executive Scott Kirby framed the decision as resistance to inserting another intermediary between United and its passengers: “To be honest, I don’t want to create another middleman.”² The stance also reflects a broader commercial and regulatory standoff. Other companies, including eBay and major banks, have restricted unauthorized agents as businesses test how much autonomy they are willing to hand software operating on behalf of customers.²

United has not ruled out a reversal. It said advancing agents may create opportunities, but account access and booking are “sensitive actions,” and protections for travelers must be established first.² For now, AI can help passengers browse; it cannot press the button that commits them to fly.