Booking.com’s AI is converting. Is your website?


Scroll to read
Booking Holdings reported its third quarter results on 3 November. The headline numbers were fine. The interesting part, if you run a hotel, was what they said about AI.
CEO Glenn Fogel called this “the most transformational time for industries since the dawn of the Industrial Revolution.” Big words. The numbers underneath them matter more, and they point somewhere independent hotels should be paying attention to.

What Booking actually said

  • AI search is taking off on Booking.com. Daily AI search visitors were up 10 times on the previous quarter.
  • AI users convert at about twice the rate. That is Priceline’s Penny and Agoda’s AI Assistant, compared with guests who don’t use them.
  • They spend more. Agoda’s AI Assistant users have an average booking value about 20% higher.
  • Outside chatbots barely register. Referrals from large language models are still “significantly less than 1%” of Booking.com’s room nights.

So the AI that is selling rooms right now is not ChatGPT. It is the AI sitting inside the app that already holds the inventory.

Whose rooms is that AI selling?

On the same call Fogel said independent hotels and alternative accommodation partners “generate the vast majority of our room nights at Booking.com.”

Put those two things together. The AI that converts twice as well is mostly selling independent hotels’ rooms, at commission. Every improvement Booking makes to its assistant makes the OTA a slightly easier place to book you than your own website.

CFO Ewout Steenbergen also said it would be “very expensive” for general AI agents to connect to 4.7 million properties directly. That is the OTA argument for the next few years: the agents will have to come through us.

Why I don’t think that’s settled

Under 1% is a starting point, not a ceiling. Booking itself said paid referrals are growing through its partnership with OpenAI. It is paying to make sure that when an assistant sends a guest somewhere, the somewhere is Booking.com.

When an assistant recommends your hotel, it still has to decide where to send the guest. If your site can’t answer the basics quickly (what the room is, what it costs tonight, what’s included, can I book it now), the guest goes to the place that can. The hotel that wins is the one that is easiest to book, not the one with the nicest homepage.

What I’d do this quarter

  1. Make your site answer like an assistant would. Plain room descriptions, real prices, availability that loads fast, policies written as text and not buried in images. If a person struggles to find it, an AI will too.
  2. Give guests a reason to book direct. Member rates, perks and a guest account they can come back to. Booking’s AI users spend more because they are guided. Your booking flow can guide too, with packages, add-ons and vouchers at the point of booking.
  3. Ask the assistants about yourself. Ask ChatGPT for the best place to stay in your area. Being mentioned is not the same as being recommended, and being recommended is not the same as being booked.
  4. Measure conversion, not traffic. A visit that doesn’t book is a guest you paid to send to an OTA.
  5. Watch your channel mix. If OTA share creeps up over the next 12 months while your demand holds, that is the AI effect showing up in your own numbers.

Summing up

Booking has told everyone exactly where it thinks this goes: AI inside its own apps, selling inventory that mostly belongs to independent hotels. Your website is also an app that holds your inventory. The question is whether it sells as well.

That is the work we do with hotels at Roomstay: a booking engine, member rates and guest accounts that make your own site the easiest place to book you.

Quotes are from Booking Holdings’ Q3 2026 earnings call, 3 November 2026.