Travel is worrying about the wrong thing. The fear doing the rounds is that AI booking agents will kill the upsell: no impulse seat upgrade, no late room upgrade, no “add breakfast for £12”. I think upsells will be fine. The thing AI agents really threaten is travel’s advertising business, and the brands paying for the top spot should be far more nervous than the brands selling extra legroom.
Here is why. A good upsell is useful information. A paid placement is just a louder voice. Agents can hear the first. They are being built to ignore the second.
What sparked this
Skift’s Rafat Ali this week called the AI agent travel’s ad blocker.
- MCR Hotels CEO Tyler Morse told the Skift Commerce Summit he wants to stop AI agents booking his hotels, because “humans are susceptible to sales pitches” and agents are not.
- Citizens analysts tested two agents, Instinct and Meta’s Muse, in a 15 November report. They found the experience “underwhelming” today, but ranked Expedia as the most exposed online travel agency long-term, then Booking Holdings, with Airbnb the most insulated.
- Expedia booked $206 million in advertising revenue in Q3 2026, according to Skift, and Airbnb’s Brian Chesky has talked up a $1 billion advertising opportunity.
- When Skift asked Instinct whether sponsored placement shaped its recommendations, it answered: “Placement isn’t a signal I use.”
That last line is the one I would pin to the wall of every OTA ad sales team.
Upsells survive. Bad upsells don’t.
The upsell numbers are huge. IdeaWorksCompany estimated $157 billion in airline ancillary revenue for 2025, 15.7% of total airline revenue, up from $67.4 billion and 9.1% in 2016. So I understand why people panic.
But look at what that money is made of. Bags, seats, meals, flexibility. Things travellers actually want. If I tell my agent “I’m 6ft 4, I need legroom on anything over three hours”, it will buy the extra legroom seat. It won’t need persuading. It needs the seat, the price and the rules in a form it can read.
What dies is the upsell that only works because a tired human is staring at a checkout page at 11pm. The pre-ticked insurance box. The “only 2 rooms left” banner. The upgrade that is mostly a nicer photo.
And agents can sell. Skift notes that Wyndham says its own AI phone agent books rooms at rates averaging 15% above published rates. That is an agent selling well because it explains value, not because it shouts. I made a similar point when Wyndham started treating AI agents as guests, and when I argued that AI agents will re-shop your hotel rooms because your pricing invited them.
Paid placement is the soft spot
Advertising is different. Sponsored listings work on one idea: whoever pays gets seen first, and being seen first wins the booking.
An agent working for the traveller does not scan. It compares. If it is any good, it compares on price, fit, reviews, flexibility and the traveller’s own history. “Who paid the OTA this month” is not on that list. Instinct said so outright.
That puts the OTAs in an awkward spot. Their ad business sells hotels and airlines a way to jump the queue. If the people doing the booking are increasingly agents that ignore the queue, what exactly are advertisers paying for? I flagged this pressure when Expedia switched on Muse before it knew the price. Citizens ranking Expedia as the most exposed fits that picture.
The fair counter-argument
Two things could prove me wrong.
First, agents are still tiny. Booking Holdings says referrals from large language models are well under 1% of its room nights. I covered that when Booking’s own AI was converting twice as well.
Second, “placement isn’t a signal I use” is a promise from one agent, today. The biggest agents will be run by Meta, OpenAI, Google and the OTAs themselves. All of them like ad money. Some will find a way to sell influence back in, maybe more quietly. That is exactly why I said Booking’s Vayo is not a neutral agent when its users all get Genius discounts.
So the honest version is this: the agent working for the traveller ignores paid placement. The agent working for a platform may sell it. Travellers will work out which is which faster than the industry expects.
What I’d do now
If you spend serious money on sponsored placement, don’t pull it tomorrow. But start shifting effort from paying for position to paying for proof.
- Make every extra machine-readable. Seat pitch, room size, what breakfast includes, cancellation terms. Agents buy what they can understand.
- Kill the dark patterns. If an upsell only works through pressure, an agent will skip it and a regulator will eventually notice it.
- Track agent traffic separately. Know how agents reach you and what they book. You can’t optimise what you lump in with “direct”.
- Ask your OTA rep a blunt question. “What does my sponsored listing do when an AI agent is searching?” If they can’t answer, that tells you something.
This also links to why banning AI agents looks like the wrong fight. Blocking agents protects the sales pitch for a while. Making your offer the obvious answer protects the business.
So here’s my question for you: if AI agents ignore who paid for the top spot, is travel’s retail media boom about to become its next stranded asset? Or will the platforms simply build agents that can be bought?
Want more takes like this? Subscribe to my LinkedIn newsletter for weekly thinking on travel marketing, AI and search.