United Airlines bans unauthorized third-party AI agents from booking and mileage redemptions

United blocks unauthorized AI booking agents

On November 11, 2026, United Airlines updated its booking, ticketing and distribution policies to prohibit AI assistants from outside parties that lack authorization from searching fares, completing bookings, changing itineraries or redeeming MileagePlus miles. According to a dev.to analysis of the change, the ban covers both United's own digital channels and the travel aggregators that resell its seats.

The airline framed the move around safety and compliance. As CBS News reported, sharing account credentials with a third-party agent opens the door to accidental purchases and unauthorized mileage redemptions. Automated check-in raises a sharper regulatory issue: FAA and Department of Transportation rules require each passenger to personally attest that their luggage contains no explosives or other banned items, something software cannot do on a traveler's behalf. There is also a commercial motive. The dev.to article notes that United chief executive Scott Kirby, speaking in Chicago that same week, said he did not want a new middleman between the airline and its customers.

Why airlines fear the automated shopper

The concern goes beyond security. Network carriers rely on their own checkout flows to sell seat upgrades, baggage add-ons and co-branded credit cards. An external browser agent that logs in, skips the merchandising screens and extracts the lowest base fare removes that ancillary revenue while the airline keeps full responsibility for supporting the customer.

Demos of agentic travel booking keep multiplying in the meantime: Meta has shown flight reservations by its Muse agent, OpenAI acquired travel startup Odessia, and Southwest Airlines plugged inventory lookup into ChatGPT. United's policy signals that carriers intend to decide which of these integrations are permitted.

Airline pricing itself rests on an assumption that agents break. Revenue management systems continuously adjust seat availability to separate price-sensitive leisure buyers from urgent business travelers, and the model works because human shoppers have limited patience, typically comparing a handful of date and airport combinations before stopping. A script has no such limits. In June 2026, the dev.to piece recounts, an insurtech founder pointed a coding agent at Etihad Airways' website and extracted 881,076 fare combinations for a single itinerary.

The infrastructure cost of exhaustive search

That combinatorial behavior lands on systems never sized for it. A November 2025 IATA distribution whitepaper cited in the article says major travel IT providers already process 3 billion to 13 billion search requests per day, with large metasearch engines reaching 100 billion, keeping conventional look-to-book ratios in the tens of thousands of searches per ticket sold. Under New Distribution Capability protocols, every query forces a live pricing engine to evaluate continuous fare curves rather than static booking classes. Industry estimates reported by PhocusWire and TravelersToday put agentic look-to-book ratios above 200,000 queries per completed purchase, at which point answering unpaid bot traffic begins to erode the margin on the seats being sold.

Suppliers are already pushing back. The dev.to analysis notes that United penalizes travel agencies that exceed look-to-book thresholds on its NDC feeds, and that Amadeus has deployed machine-learning filters that cut more than 70 percent of unproductive search traffic on Air France-KLM's network while testing precomputed fare caches. Caching reduces load, but dynamic prices can shift between a cached quote and live checkout, producing failed bookings.

What the ban implies for agent design

For developers of consumer-facing agents, the article draws three practical lessons. First, raw browser automation with delegated credentials is fragile, because airlines and financial platforms treat unapproved automated sessions as policy violations and block them. Second, search tools need explicit query budgets and stopping rules rather than open-ended permutation loops that trigger rate limits and agency penalties. Third, any step involving regulated attestations, irreversible payments or loyalty balances should pause at an explicit user confirmation gate instead of automating the final submit. Comparing itineraries on public schedules remains viable; executing a booking requires authorized API contracts and clear liability boundaries.

Why it matters

United is among the first large consumer services to formalize where AI agents may not operate, and the policy effectively splits agent development into two tracks: read-only exploration, which stays possible, and transactional execution, which will require formal partnerships and APIs. Developers building agents for travel, or for any regulated and margin-sensitive marketplace, should assume browser automation will not survive contact with production, and instead design for query limits, human confirmation at irreversible steps and official integrations from the outset.