United Restricts AI Bookings on Flights as Hotels Debate Room Access
United now restricts unauthorized AI agents from booking flights. MCR's Tyler Morse wants similar hotel controls, citing agents that skip upsells, while one column says guests will choose.

By Marcus Bennett
United now restricts unauthorized AI agents from booking its flights, while MCR's Tyler Morse wants a similar block on hotel reservations. The contrast has exposed a practical dispute over who controls the booking, the upsell and the customer relationship.
What changed at United?
United's restriction applies to agents that lack authorization to book flights. The restriction does not define authorization, identify affected booking systems or disclose an enforcement date. It also provides no volume of blocked transactions. Hotels therefore have a warning signal, not a proven operating model.
In the same week, Morse said he would like to block AI agents from his hotels because agents do not take upsells. He did not announce a rollout, list affected properties or quantify the revenue at stake. The column does not specify whether his preference covers company-operated or franchised hotels.
Morse's position places hotel economics ahead of guest convenience, although the proposal remains a stated preference rather than an operating decision.
Why marketplaces may resist
Benedict Evans posed the commercial risk in a new deck: “Do you want to be someone else’s dumb API call?” A marketplace earns its place by gathering demand, organizing supply, building trust and encouraging repeat business. An agent placed above that service can turn the marketplace into plumbing, with its interface serving as the product.
Hotels retain an advantage because their product cannot be copied into software. A hotel remains a building with rooms, beds and showers, and no guest sleeps in a digital bed. That physical inventory gives hotels leverage if an intermediary tries to own the entire customer relationship.
Why owners may lose the argument
Martin Soler, writing in his weekly Tell newsletter on hotels, technology and marketing, argues that hotels overestimate how much guests care about a preferred process. Luxury hotels once resisted website booking engines because telephone reservations felt more elegant and on-brand. Guests wanted to book online, so the hotels eventually allowed it.
Some hotels also believed they could screen for the right kind of guest. That effort faded, Soler wrote, because money is surprisingly democratic. His practical conclusion is direct: “The hotel’s job isn’t to protect the old way.” Guests will use an agent if they want one, and an expensive hero image will not compel them to use a hotel’s own booking engine.
The same pattern appeared with mobile. Hotels treated the channel as too small and unimportant until online travel agencies had done the promotional work and secured the best placement.
Agents remain at that early stage. The column cites a tiny share of hotel search activity, although much consists of demonstrations still looking for a business model. Soler argues that description does not settle the case because all new technology starts without a settled model.
What would agents replace?
At present, an AI agent would replace some manual room comparison and form completion with automated booking. The account identifies no hotel already using the technology at scale. It also does not say who pays, whether agents carry booking fees, or how hotels would allocate support and risk.
Soler argues that most hotel booking flows are tolerated rather than liked. Guests do not wake up excited to compare room types. They want confidence, convenience and a fair price, and an agent that delivers those conditions faster will gain use. The hotel website remains a direct channel, but its inconvenience gives agents a plausible opening.
The risks resemble those hotels already faced with online travel agencies, metasearch services and mobile booking. Hotels have worked through versions of those problems, imperfectly, and survived. Owners still control their inventory, systems and risk. Guests still decide where to spend their money, a choice that has favored guests in earlier distribution disputes.
What remains unproven?
The column offers no booking-share forecast, conversion rate, fee schedule, implementation cost or hotel-level margin impact. It also reports no measured difference in upsell revenue between agent and direct bookings. Those omissions leave the central business case directional rather than financial.
For hotels, the decision is whether an automated intermediary adds enough convenience and demand to justify surrendering some direct-booking control. The next phase should be judged by actual agent transaction volume, measured upsell conversion and disclosed fees, rather than another polished demonstration of a completed reservation.
More from Marcus Bennett
Show full bio
Market editor covering technology and product news in Restaurant and hotel operations at The Pass Brief.
798 articles

