Airbnb this week launched an AI-powered search as part of its fall update, marking a significant step for a company that has been deliberate about adopting artificial intelligence. In an interview with TechCrunch, CEO Brian Chesky reiterated his view that chatbot-style interfaces are not the right fit for e-commerce or travel planning.
"A chatbot isn't the right interface for e-commerce," Chesky said. He noted that chatbots present only a few options at a time and require multiple conversational turns to reach a result, making them inefficient for browsing. He added that travel studies show people often derive more pleasure from planning and anticipating a trip than from the trip itself, so the interface should preserve inspiration and discovery rather than reduce everything to quick transactions.
Chesky identified a second problem: chatbots are single-user tools, while travel planning is often a group activity involving family or friends. Over the next three to six months, Airbnb plans to explore interfaces that enable "multiplayer" AI, allowing several people to use the same AI assistant simultaneously when planning a trip together.
He does not see the current version of Airbnb's AI search as the final form. "There's going to be some interface that is [...] something between a chatbot and what you see in the first version we shipped," he said. Chesky noted that startups like Monogram and Wabi are experimenting with "generative" interfaces built on the fly by AI, and he anticipates a mix of deterministic (pre-designed) and generative interfaces emerging in the future.
On the topic of AI agents that carry out tasks on a user's behalf, Chesky said Airbnb wants to make itself more "agent-friendly" by building robust software infrastructure. He sees agents as potential lead generators for the platform, but cautioned against a future where apps become mere data layers beneath a universal interface. "There is a reason Airbnb's app is a different design than DoorDash, Shopify, and Amazon," he said. "Our interfaces should be adaptive to the utility and the function that we're doing."