Booking Holdings reported its third quarter results on 3 November 2026, and for the first time the company put hard numbers on how its AI assistants perform. The CEO and CFO prepared remarks and the earnings call covered conversion, booking value, AI search traffic and how few room nights large language models actually send. If you sell booking engines, channel managers, revenue tools or guest messaging to hotels, these figures will turn up in your prospects' heads before they turn up in your demos.

Below are seven numbers from the call, what each one says, and what it means for how hotel tech vendors sell, price and position over the next few quarters.

1. AI assistant users converted at about twice the rate of non-users

CFO Ewout Steenbergen said users of Priceline's Penny and Agoda's AI Assistant converted at approximately twice the rate of non-users, according to MarketBeat and Seeking Alpha. Agoda's assistant is now live across 21 markets and moves travellers from an idea to a bookable trip using Agoda's live inventory, per the prepared remarks.

What it means for vendors: the OTA now has a conversion story built on conversation, not just on price and reviews. If your pitch to hotels is that direct booking converts better, expect a revenue manager to ask how your booking engine handles a guest who wants to ask questions before paying. Have an answer: a live chat or assistant on the booking path, or a clear integration with one. Update the objection notes in your booking engine objection playbook before the question comes up.

2. Agoda AI Assistant users booked about 20% more per booking

Steenbergen said Agoda AI Assistant users had average booking values about 20% higher than non-users (MarketBeat). Booking described this as one of three early indicators alongside engagement and conversion.

What it means for vendors: booking value, not just booking count, is now part of the AI argument. Vendors selling upsell, packaging or guest messaging tools should measure their own effect on average booking value and publish it in case studies. A claim like "our assistant lifts booking value" without a figure will now be compared with a number the largest OTA has already said out loud.

3. Daily AI search visitors rose tenfold quarter on quarter

Booking.com's daily AI search visitors using trip search increased tenfold in the third quarter compared with the second, according to MarketBeat. The prepared remarks also describe agentic tools that let customers cancel a booking or get answers about their stay without contacting customer service.

What it means for vendors: guests are getting used to asking software questions in plain language. Hotel staff will expect the same from the tools they buy. If your product still depends on menus and reports, the demo needs to show where a hotelier can ask and get an answer, or you need a clear position on why that is not your job.

4. LLM referrals are still well under 1% of room nights

This is the number that should calm the hype. Large language model referrals account for significantly less than 1% of Booking.com room nights, despite accelerating in recent weeks (MarketBeat, Seeking Alpha). Skift reports that a Booking.com survey earlier this year found 64% of travellers in Asia Pacific use AI to plan trips, and Booking.com's Laura Houldsworth told Skift that travellers move between several chatbots and Google before coming to Booking directly.

What it means for vendors: do not sell "AI booking" as a revenue channel hotels must buy now. Research happens in AI; the booking still happens on a site or app that has live rates and a checkout. The honest pitch for most vendors is readiness: clean content, accurate rates and availability, and structured data that AI tools can read. We covered the visibility side in a 10-question AI answer self-assessment and the agentic side in Google's AI hotel booking rules.

5. The ten largest chains are only a low double-digit share of room nights

CEO Glenn Fogel said independent hotels and alternative accommodation partners generate the vast majority of Booking.com room nights, and that the ten largest global hotel chains account for only a low double-digit percentage, according to the prepared remarks. Alternative accommodation alone was about 37% of Booking.com room nights (MarketBeat).

What it means for vendors: Booking's core supplier is the same small independent many hotel tech vendors target. Its pitch to that hotel is reach without upfront marketing cost: partners pay only when they earn revenue for a room. That is a pricing model, and your prospect is comparing your subscription with it. If you sell to independents, show the cost of your tool against the commission it helps the hotel avoid, using real figures. Our OTA commission sales argument and the Hotel Software Pricing Index are useful starting points.

6. Direct B2C mix stayed in the mid-60% range

Booking's direct B2C mix remained in the mid-60% range over the past four quarters, the mobile app share of room nights was in the high-50% range, and higher-tier Genius members accounted for a high-50% share of Booking.com room nights, per MarketBeat. Fogel said partners fund much of Genius and that the company plans to improve it, possibly with benefits that have low marginal cost for partners.

What it means for vendors: the OTA is building its own direct relationship and loyalty layer with the guest, funded in part by hotel discounts. Vendors selling CRM, loyalty or direct booking tools have a clear story here, but it has to be about the hotel's own repeat guests, not a promise to beat Booking at acquisition. Position against the discount cost, not against the OTA.

7. More than 100 payment methods in over 50 currencies

Fogel listed the services Booking provides partners: payment processing for more than 100 payment methods in over 50 currencies, customer service in more than 40 languages with AI agents and people around the clock, and on-the-ground Partner Services teams (prepared remarks). Merchant gross bookings were about 75% of the total (Seeking Alpha).

What it means for vendors: when an independent hotel weighs your product against "just using Booking", it is weighing a bundle. Payments, multilingual support and local help are part of what the commission buys. If your product replaces any of that, say so plainly and price it as a replacement. If it does not, do not imply it does.

The numbers at a glance

What else is moving

Booking said it works with Meta, OpenAI, Google, Anthropic and Amazon, and that OpenTable's native OpenAI integration generated 190,000 seated covers in September (MarketBeat). PhocusWire reports that Agoda is the first digital travel platform to connect its AI travel assistant to Meta's Muse agent, live in Singapore alongside Trip.com and Klook. Booking also launched two AI-native products, Lola and Vayo, which it calls testing grounds (prepared remarks). For Q4, Booking guided room nights up 1% to 3% and flat adjusted EBITDA as it increases investment in AI and technology infrastructure (Seeking Alpha).

The pattern is that the largest intermediaries are connecting to every AI surface at once. Hotel tech vendors cannot match that breadth, so pick the one or two connections your customers actually need and make them work well. It is the same lesson as Expedia's AI property answers: the platform with clean, structured data gets used.

What to do this quarter

• Add an answer to "OTA assistants convert at twice the rate" to your booking engine and direct booking talk tracks.

• Measure your product's effect on booking value or conversion and put the figure in one case study.

• Remove any claim that AI chat is a large booking channel today; sell data readiness instead.

• Build a one-page comparison of your annual cost against the commission it helps an independent hotel avoid.

• Choose the AI or agent integrations your target hotels ask for, and document them.

These choices sit in the position and price layers of a hotel tech go-to-market strategy. If you want to check whether your current pitch still holds up against the OTA's new numbers, start with a free GTM leak audit.

Frequently Asked Questions