AI agents are replacing OTAs. Hotels must own the guest relationship.
For 20 years, OTAs sat between hotels and guests. Booking, Expedia, Hotels.com took 18% margin in exchange for visibility and traffic. It was a bad deal — but the only one.
What's changing
AI agents (ChatGPT, Perplexity, Claude) are becoming the new "choice layer". The guest no longer opens Booking.com. They ask: "find me a mountain resort in March, no kids' animation, great spa". The AI recommends. The guest books directly with the hotel. The OTA disappears from the chain.
Sounds like a win over OTAs. But it's a new problem: AI only recommends hotels it "knows". And "knows" means hotels with rich relationship data, reviews, loyal audience, and a history of direct bookings.
Who disappears
A hotel that relied on OTAs for 20 years has none of:
- guest profile data beyond "name + email from reservation"
- history of direct relationships
- loyalty programme that works as memory
- behavioural segmentation
When an AI agent checks "does this hotel have a sustainable audience and good reputation" — that hotel is invisible. It's happening now: the first AI travel agents rank by data depth, not PR budget.
Window: 2-3 years
By end of 2028, AI agents will be the default travel-search method for 30-40% of audience (per Skift, FT, and Booking Holdings' own Q4 2025 acknowledgement of the threat).
In these 2-3 years, hotels must build retention infrastructure — a layer that turns the first stay into a long-term relationship: persistent guest profile, multi-touch engagement, between-season reactivation, and making that relationship visible to AI agents through standardised data.
What to do next week
1. Audit: what guest data do you have today beyond your PMS? (typically — almost nothing)
2. Identify who didn't return in the last 12 months. That's your sleeping pool.
3. Decide: continue paying OTA commission, or build direct-channel infrastructure that pays back in one season.
The window is open. But not for 10 years. For 2-3.