Airbnb's AI customer-service assistant now resolves nearly 45% of user inquiries without a human agent, up from about a third at the end of last year, CEO Brian Chesky said. The assistant operates in more than 50 languages, with voice-call support planned later this year. For customer support teams, the number matters less than the sequence: Airbnb deliberately automated support before touching search, because Chesky saw customer service as the hardest AI problem to solve.
"The reason we decided to start with this is we want to focus on the hardest problem in AI, which we thought was customer service," Chesky told analysts.
AI moves into search and booking
Airbnb is now pushing AI into the parts of the platform travelers see before they need help. The company is testing an AI search toggle that lets users type natural-language queries instead of relying on traditional filters. Results include AI-generated listing titles and highlights that read more like a conversation than a search results page.
"The titles could actually be AI generated and they can be conversational as if you're reading a chatbot, but more visual," Chesky said.
Later this year, Airbnb plans to launch AI-powered home comparison, letting guests compare listings side by side before booking. The company is also applying AI to host-facing tools, including faster listing onboarding.
AI behind the scenes
Some of Airbnb's AI never surfaces to a traveler. The company uses machine learning-powered "anti-party technology" to flag bookings at higher risk of becoming disruptive, unauthorized parties. The system analyzes listing type, stay duration, travel distance, and last-minute timing, and only about 0.15% of global bookings get flagged as a potential risk, according to Airbnb.
That same logic extends into product development. AI has cut the time from concept to feature launch by as much as 60%, and AI now writes roughly 60% of the company's new code, Chesky said. He has described that shift not as replacing engineers, but as letting the same team ship substantially more in the same amount of time.
Booking Holdings makes a different bet
Airbnb's rival Booking Holdings is arguing that the platform still holds the real value in travel. Generative AI can recommend a hotel, assemble an itinerary, or suggest a cheaper date, but it cannot independently guarantee a room, process a payment, or resolve a dispute, said Daniel Marovitz, Booking Holdings' chief technology officer.
The two companies are converging on the same technology from opposite starting points. Airbnb is betting that owning the AI experience at every step - support, search, discovery, safety - keeps the traveler inside its ecosystem. Booking is betting that even if an AI assistant plans the trip, the platform that completes the transaction is where the real value and risk still live.
Why this matters for customer support professionals
Airbnb's trajectory shows a support automation model that works: solve the highest-volume, most repetitive inquiries first, at scale, across languages, before layering AI onto other parts of the product. The 45% resolution rate did not come from deploying a chatbot and hoping for the best. It came from treating support as the hardest AI problem and building from there. For support teams, the lesson is to measure resolution rates the way Airbnb does - as a percentage of all inquiries, not just the ones routed to automation - and to expect that internal AI tools that speed up engineering will change how quickly new support features ship. If you are building an AI support strategy, the AI Learning Path for User Support Specialists covers similar ground, and AI for Customer Support Courses offers training built around these workflows.
Your membership also unlocks:
