Dorian Destinations

Latest Travel News

What The Reporting Actually Says

AI Travel Agents: One Real Policy and Three Opinions


Franco Interian, Co-Founder and Travel Advisor at Dorian Destinations

By Franco Interian

Co-Founder & Travel Advisor, Dorian Destinations · Published November 12, 2026 · ET

Travellers moving through a sunlit airport terminal, the transaction that AI travel agents are now arguing over
An airport terminal. The photographer states no location and we are not going to invent one. Photo by osamu nakazawa on Unsplash.

Have a human build your next trip →

Four stories about AI travel agents landed on Skift in a single day, 11 November. Read the headlines and the industry is slamming the door. Read the articles and almost nothing has actually been decided.

Exactly one of the four is a real company policy. United Airlines said on Wednesday it is taking steps to restrict third-party personal AI agents from its digital platforms, and from reaching it through online travel agencies and aggregators. The other three are a hotel executive’s stated wish, a bank’s proposal about who should pay when things go wrong, and a payments partnership that the headline framed as an AI story more firmly than the article did.

We are writing this up because the gap between those headlines and that reporting is the story. If you are a traveller wondering whether AI travel agents can still book your flight, or an advisor wondering whether this changes your business, the honest answer today is: the industry is arguing, and it has not finished.

Everything below comes from Skift‘s reporting of 11 November 2026, read the following day. Where a figure appears only in Skift’s auto-generated key-points box rather than in the reported body, we say so, because those are not the same thing. Where an article does not state something, we say it does not state it rather than guessing.

And our own position, stated plainly: we are travel advisors. We have an obvious interest in a story about AI travel agents replacing human advisors, so treat our reading accordingly, and note that we have not predicted the outcome below. We do not know it.

“Everyone is going to ban the agents because it’s in their interest to do so.” — Tyler Morse, chairman and chief executive of MCR Hotels, speaking at the Skift Commerce Summit

What United Airlines actually said about AI travel agents

United Airlines is restricting third-party personal AI agents, and it is the only one of the four with a stated company position rather than an opinion. Skift reports the airline said on Wednesday that it is taking steps to restrict access, with the aim of prohibiting use by unauthorised third-party personal AI agents or assistants.

The scope is wider than United’s own app and website. The restriction is described as reaching the airline indirectly too, through online travel agencies and aggregators.

The reason United gives is security: sharing account credentials with a third-party agent can expose sensitive information. That is a narrower argument than “we do not want machines buying our seats,” and it is worth noticing which argument a company chooses.

What the article does not say, and we are not going to fill in: there is no effective date, no statement of whether the restriction is live or planned, and no description of the technical method. Skift’s headline says United “moves to block.” The body says United is taking steps to restrict. Those are not the same claim, and the difference matters if you are deciding whether something changes for you this week.

What the hotel side actually said about AI travel agents

Tyler Morse wants to stop personal AI agents booking rooms at MCR Hotels, and says he cannot yet work out how. Skift reports the MCR chairman and chief executive making that case at the Skift Commerce Summit on Wednesday, alongside the key-points line that he “can’t figure out how to do it.”

His reason is commercial rather than technical, and he is refreshingly direct about it. An AI assistant makes more rational decisions, and it cannot be upsold the way a human guest standing at a desk can. Skift notes that this is a different argument from United’s: the airline cited data privacy and regulatory risk, while the hotelier is talking about margin.

Two further numbers appear in Skift’s key-points box rather than the reported body, so treat them as second-hand: that Skift found ten of fifty hotel brand sites explicitly addressing AI crawlers in robots.txt, and that Skift Research put global hotel distribution costs at roughly $105 billion in 2023, about $75 billion of that indirect. We have not independently verified either and we are flagging where they sit.

What the article does not say: any mechanism, any date, or whether MCR has done anything at all yet. This is an intention, not a policy, and reading it as a ban on AI travel agents would be reading past the reporting.

A traveller on the phone in an airport waiting area, the kind of human conversation AI travel agents are being built to replace
A traveller on the phone at the gate. The photographer titles it simply “Waiting…” and states no location. Photo by David Geneugelijk on Unsplash.

Who pays when AI travel agents book the wrong trip?

Citi’s position is that liability should sit with the company behind the agent, not the hotel or the airline. Abhinav Anand, Citi’s head of value cards, lending and commerce for US consumer cards, made the argument at the same summit on Wednesday: responsibility belongs with the agent itself and the company that brought it.

Skift reports one route being discussed, that agents transacting over Visa or Mastercard might have to accept some liability. It also frames the trade-off honestly — putting liability on the AI company would spare hotels and airlines, but could slow down how fast agents are able to transact at all.

This is the question that actually decides the future of AI travel agents, and it is the one getting the least attention. A machine that books your holiday is interesting. A machine that books the wrong holiday and leaves an unresolved question about who refunds you is a problem nobody has solved. Skift describes banks, payment networks, merchants and AI companies all still working out the answer.

What the article does not say: any date, any adopted rule, and nothing specific about what this means for travel agencies. It is a proposal.

The plumbing story nobody led with

Duffel and Katanox announced a partnership on Wednesday aimed at the plumbing beneath AI travel agents, connecting hotel distribution and payments — and despite the headline, the reported body describes no AI-agent feature at all.

The problem it targets is unglamorous and real: travellers arriving at hotels that cannot find their booking, and sellers chasing commissions months after checkout. Duffel’s reach is given as 1.6 million hotel properties worldwide, the single hard figure across all four stories’ body text.

The AI framing comes from the lede and from Skift’s key-points box, which states the setup was built with AI booking agents in mind. Named in the box, not the body: Steve Domin of Duffel and Paul Beukers of Katanox. Skift’s own summary notes the two chief executives describe settlement timing differently and have not reconciled it, which is the sort of detail that tells you how early all of this is.

What the article does not say: an effective date, or any particular capability aimed at AI travel agents beyond the framing.

Story What it actually is Status Date given
United Airlines Restriction on third-party personal AI agents Stated company policy None
MCR Hotels Executive wants to block AI bots Intention; method unsolved None
Citi Liability should sit with the AI company Proposal None
Duffel & Katanox Hotel distribution and payments partnership Announced, future tense None

All four from Skift, 11 November 2026. The status column is our characterisation of what each article’s body supports, not Skift’s wording. Not one of the four carries an effective date.

Is this a crackdown on AI travel agents?

No — it is a debate, and calling it a crackdown would be reading the headlines instead of the reporting. Two of the four stories concern restriction, and only one of those is an actual airline policy. The third is about who pays. The fourth is about payment plumbing that happens to be useful to agents.

Two of the four companies here are arguing for the opposite outcome on AI travel agents. Citi is working out how agent transactions can be made safe enough to proceed; Duffel and Katanox are building infrastructure that makes them work better. That is not an industry closing ranks.

What is genuinely true across all four: nobody has published a date, nobody has described a working mechanism, and the only named hotelier who wants to block AI travel agents says in the same breath that he does not know how.

Travellers photographing aircraft through a terminal window, the ordinary moment AI travel agents never see
An airport terminal in Shanghai, as the photographer states — worth naming, since none of the policies discussed here are Chinese. Photo by Margo Evardson on Unsplash.

What AI travel agents change for you this week

Nothing is live that you need to act on today, and anyone telling you otherwise is selling something. But three things are worth holding onto.

Do not hand your airline login to anything. United’s stated reason was credential sharing exposing sensitive information, and that reason is sound whoever is making it. AI travel agents that need your frequent flyer password to work are a risk regardless of which company blocks them first.

Ask who carries the booking. The Citi question is the one that matters to a real traveller: if something goes wrong, who fixes it. When we book, the answer is us — a named person with the supplier relationship and the ability to pick up a phone. That is the whole argument for a human advisor, and it is the one part of this story that is not in dispute.

Watch the liability answer, not the blocking announcements. Blocks get headlines and quietly lapse. The rule about who pays when an AI travel agent books the wrong week in the wrong city will decide what these tools are actually allowed to do.

More from Dorian Our Travel Specialists — the people who answer when a booking goes wrong, which is the function this entire debate is circling.

AI travel agents FAQ

Has United Airlines blocked AI travel agents?
Not exactly. Skift reported on 11 November 2026 that United said it is taking steps to restrict access by unauthorised third-party personal AI agents, across its own platforms and indirectly through online travel agencies and aggregators. The article gives no effective date and does not say whether the restriction is already live.

Why do airlines and hotels want to restrict AI travel agents?
For two different reasons. United cited security — sharing account credentials with a third-party agent can expose sensitive information. Tyler Morse of MCR Hotels gave a commercial reason instead: AI assistants make more rational decisions and cannot be upsold the way a human guest can.

Who is liable if an AI agent books the wrong trip?
Unresolved. Citi’s position, stated at the Skift Commerce Summit on 11 November 2026, is that liability should sit with the agent and the company behind it rather than the hotel or airline. One route discussed is that agents transacting over Visa or Mastercard might have to accept some liability. No rule has been adopted.

Can AI travel agents still book flights and hotels right now?
Broadly yes. Of the four stories, only United has stated a restriction, and even that carries no published date or described mechanism. MCR’s chief executive says he wants to block them and has not worked out how.

Does this mean human travel advisors are safe?
It means nothing about AI travel agents has been settled. The useful read is narrower: the open question in all this is who takes responsibility when a booking goes wrong, and a named human advisor is an answer to that question that already exists.

The verdict

Four stories in one day looked like a wave and turned out to be a conversation. One airline has stated a restriction without a date. One hotelier wants a block he cannot build. One bank wants the liability moved. One partnership is quietly making agent transactions work better, which rather undercuts the idea that everyone is pulling in the same direction.

The reason to pay attention to AI travel agents is not that they are about to be banned. It is that the industry has not yet decided who is accountable when one of them gets your trip wrong — and until it does, the thing worth paying for is somebody whose name is on the booking.

One real policy, no dates, and nobody who knows how to enforce it yet.

Sources: all four stories are Skift, 11 November 2026 — United Airlines, MCR Hotels, Citi and Duffel and Katanox, all read on 12 November 2026. Figures and names drawn from Skift’s auto-generated key-points boxes rather than the reported body are labelled as such in the text above, and we have not independently verified them. Where an article states no date or no mechanism, we have said so rather than inferring one. The status column in the table and the advice in this piece are ours.

Talk to a Dorian advisor →

Travel News · Flights · Booking · Travel Advisors · Corporate Travel

Scroll to Top