Airbnb maps a path from AI search to interoperable travel agents
Airbnb has rolled out AI-powered search in its fall update and is now planning a broader shift toward agent-ready travel experiences. In an interview, co-founder and CEO Brian Chesky said the company will spend the next three to six months exploring “multiplayer” AI interfaces for people planning trips together. He also said Airbnb intends to introduce voice agents for search and customer service this fall.
Chesky’s position is that neither a text chatbot nor Airbnb’s newly released search experience represents the endpoint for travel discovery. He expects an interface somewhere between conversational AI and the company’s current product, one that can be fast for straightforward requests without taking away the browsing and inspiration involved in planning a trip.
Travel planning needs more than a chatbot
Chesky said chatbots are not well suited to browsing or shopping because they present only a limited set of choices at a time and require repeated turns before a user reaches a result. That limitation is particularly relevant to Airbnb, where travellers may want to inspect options, compare properties and use the planning process itself as part of the experience.
He also identified collaboration as a missing feature in conventional chatbot interactions. Travel decisions often involve family members or friends, while most chat interfaces are designed around a single user. Airbnb’s proposed multiplayer work is intended to address that gap by allowing several people to use AI within the same planning process.
The company expects both deterministic interfaces, designed in advance with predictable controls, and generative interfaces that AI can assemble dynamically. Chesky does not believe consumer software will become entirely generative, arguing that designers can create better software experiences than a consumer can prompt from scratch.
Making apps usable by agents
Airbnb is also preparing for consumer agents that can act on a user’s behalf. Chesky said these systems will require robust infrastructure from service providers, and that Airbnb wants to make its app more agent-friendly. He described agents as potentially useful lead generators rather than a reason for apps to disappear into a universal, simplified interface.
For Airbnb, an agent still has to coexist with product capabilities including rich browsing, host messaging, comparisons, identity verification, maps and adding further items to a trip. Chesky said an external agent may need to hand off into the service, or its developer would need a more capable software development kit that can bring those controls into the agent experience.
He envisaged an Airbnb agent in the Explore tab, another in customer service and agents in other service areas, potentially followed by a broader Airbnb agent. Chesky said such systems could eventually interoperate with other agents through MCP, the standard designed to let AI agents connect with other software.
An operating-system gap remains
Chesky argued that the larger industry constraint is the absence of a true operating system for AI. In his view, AI applications built on iOS, macOS and Windows are not themselves AI operating systems, because agents cannot yet interact at the operating system’s kernel level through a full developer platform.
He said the current contest to become the primary agent through which users access everything overlooks that developer-platform problem. Chesky cited ChatGPT’s early app-store effort as an example of why an app marketplace needs both a software development kit and an operating system model if third-party functionality is to work fully.
His experience using Airbnb through Muse and Instinct reinforced that concern: he said those interactions do not work well, with Airbnb listings performing worse than hotels. Chesky’s conclusion was that consumer AI requires progress at a lower technical layer, alongside interoperable agents and richer user interfaces.
Apple and Google remain central to any more fundamental move from apps to agents, Chesky said, because they control the dominant platforms. Until a new platform emerges, businesses should plan for agents inside applications while retaining the specialised controls customers need to browse, collaborate, verify details and complete transactions.