Banning AI agents from booking travel is not a safety policy. It is a bet that your customers will stay easy to upsell. And it is a bet travel brands will lose.
Yesterday two big names said out loud what a lot of the industry has been thinking. United Airlines said it is restricting access for unauthorised third-party personal AI agents, both on its own website and app and indirectly through travel agencies and aggregators. On the same day, at the Skift Commerce Summit in New York, MCR Hotels chairman and CEO Tyler Morse said he wants to stop AI agents booking his hotels, and that “everyone is going to ban the agents because it’s in their interest to do so.”
The difference between the two is the interesting part. United talks about security. Morse was honest.
United’s reasons are real. They are not the whole story.
United’s case is reasonable on the surface. It says handing your MileagePlus login to a personal AI assistant can expose sensitive information and lead to unintended purchases, account changes or miles being spent without your approval. It also points out that some steps, like check-in, need a human to confirm that bags don’t contain banned items. According to PhocusWire, the policy covers bots, browser extensions and browser features acting for a customer without United’s permission, and travel agencies can only use their own AI tools with United’s say-so.
Read United’s own help page and the shape of the policy is clear. You can still use AI to research and plan. You just have to come back to united.com or the United app to buy. Does United work with any third-party agents? “Not yet.”
So the AI can do the thinking, and United keeps the moment of purchase. That moment is where the seat upgrades, the bag fees, the card offers and the loyalty nudges live. Security is a genuine concern. It is also a very convenient one.
Morse said the quiet part
Morse did not bother with the security framing. As Skift reported, his objection is that AI assistants “make more rational decisions”, and travel merchandising is built to trigger impulse purchases. A person can be talked into the bigger room. Software, on the whole, can’t.
I respect the honesty. But think about what that argument actually says to a traveller: “We prefer you when you are easier to persuade.” That is not a strategy you can say on a billboard. It is a strategy that only works while customers don’t have a better option.
And they are getting better options quickly. We have already seen AI agents booking flights through an OTA you never searched, OpenAI buying a travel agent because ChatGPT wants the booking, and agents set up to re-shop hotel rooms when the price drops. The demand is not going away because one airline updated its distribution policy.
Blocking hands the relationship to someone else
Here is the risk. If your brand is the awkward one that the traveller’s AI can’t book, the AI will not give up. It will recommend the one it can book. That might be your competitor. It might be an OTA with a cleaner integration. Either way, you have trained the assistant to route around you.
Airlines have more power than most here, because a route is a route. Hotels have far less. If an independent or mid-scale hotel blocks agents while the place across the road welcomes them, the agent doesn’t feel loyal. It just moves on. I have argued before that Wyndham is smarter to treat AI agents as guests than to hide behind loyalty, and this week makes that point stronger, not weaker.
The better answer was in the same room
Also at the Skift Commerce Summit, Citi said AI companies should pay when their agents book the wrong travel. Abhinav Anand, Citi’s head of value cards, lending and commerce, called “who takes the agent liability” the biggest open question for banks, payment networks, merchants and AI firms.
That is the grown-up conversation. Not “ban the agents”, but “set the rules”: which agents are authorised, what they can do, who pays when they get it wrong, and what data you share with them. United even hints at this, saying it wants “standardised, secure technology integrations” with accountability in place. Fine. Then build them, publish the terms, and say when.
Because a ban with no roadmap looks less like caution and more like stalling. And stalling is exactly how travel lost control of distribution last time, first to the OTAs and then to Google. Travellers already told us they will let AI plan the trip but hesitate to let it pay. That gap will close. The brands that have trusted, authorised agent access ready when it does will win the booking.
What travel marketers should do now
- Decide your agent policy on purpose. Block, allow or authorise, but write it down. Don’t let your bot protection vendor decide it for you. Start with what your robots.txt is already saying to AI crawlers.
- Win the research stage anyway. Even United lets AI plan. If the assistant recommends you, the customer will come and finish the booking. That is a GEO job, and it is yours to win or lose.
- Sell value, not friction. If your revenue depends on customers not noticing what they are buying, AI is going to find that out. Better to fix the offer than ban the shopper.
Morse may be right that plenty of brands will try to ban the agents. I just don’t think the ones that do will be the ones writing the rules in three years. What do you think: is United protecting customers, or protecting the upsell?
Want more takes like this? Subscribe to my LinkedIn newsletter for weekly travel marketing, AI and search insights.