Booking's Under-1% Share Is Not a Conversion Rate

If you read that AI traffic doesn't convert because Booking.com gets almost nothing from it, you read a number that can't carry that claim.

Booking Holdings told investors on its November 3, 2026 earnings call that referrals from AI chatbots have accelerated but are still "significantly below 1%" of Booking.com room nights. That is a share of booked room nights. A conversion rate divides bookings by visits, a different denominator, and the call gives no such rate for chatbot referrals.

My read is that a channel can be tiny and still convert well. The call offers no proof for chatbot referrals either way.

The stat spread on X within days. The earnings-call quote account The Transcript posted the line on November 7, and Kevin Indig, who writes the Growth Memo, responded to it with "Seeing the same weak conversions from AI and agents across the board." Replies treated Booking's number as the proof. His post does not say what data sits behind the claim, and the Booking figure cannot supply it.

An analyst asked how newer agents compare with the chatbot referrals the company already sees. The TradingKey transcript credits the answer to Ewout Steenbergen, one of the two Booking Holdings speakers on the call: "Yes, we have seen some acceleration in the recent weeks, but it's still significantly below 1%. I also would like to point out that the vast majority of that traffic is coming to us and that is based on third-party data sources. "

That answer is about how large the channel is and how much of the available referral traffic Booking captures. It is not about what happens after a visitor lands.

In a different answer, Steenbergen listed four areas where Booking sees early gains from AI. On conversion, he said Priceline's Penny assistant and Agoda's AI assistant users "have converted at approximately twice the rate of nonusers." The transcript gives no sample size, time period or definition of conversion for that comparison. I treat it as a company claim and not a study.

It also describes a different group of people. These are users of assistants built into Booking's own brands. They are not strangers who arrived from an answer in ChatGPT. Someone who opens an assistant inside Agoda is probably already planning a trip, which would lift conversion there on its own. It says little about whether outside chatbots send buyers.

Booking Holdings CEO Glenn Fogel hedged more openly. Asked whether AI tools create new demand, he said "the referral numbers are pretty darn small, and we believe we're getting a good number of the total number out there." He added: "So perhaps we end up with things that are actually going to convert at a higher rate." The word that carries that sentence is "perhaps." It is a hope about the future, and he gave no conversion rate for chatbot referrals on the call.

Retail offers a second data point that points the other way. Adobe's August figures, as reported by The Agentic Operator, say AI-referred shoppers on US retail sites converted at a 48 percent higher rate than other visitors. That is a different industry and a secondhand report, and it can't settle the Booking question either.

Indig's side has one argument I take seriously. Travel is a long, expensive purchase. If people ask ChatGPT about a hotel and then search Booking.com by name, the visit lands in direct or branded search, and a last-click report credits something else. Weak measured conversion from AI could mean weak conversion, or it could mean the credit went elsewhere. Nobody has published the numbers that would split those two.

I would not draw a conclusion from any of these. Booking published a share of room nights and, as far as the call transcript shows, no conversion rate for outside chatbots. Until someone shows both numbers for the same surface, "weak conversions" is a hunch that I can't rule out and can't prove. My guess is that the answer will differ a lot by surface, because an assistant user inside Agoda and a stranger arriving from a chatbot answer are not the same buyer.