Airbnb is turning into an AI-native travel guide

While everyone else rushed to slap chat interfaces onto search bars over the past two years, Airbnb took its time. Now, CEO Brian Chesky is ready to show his cards.

During the company’s Q4 earnings call, Chesky laid out plans to build a truly AI-native platform. The goal is an app that doesn’t just show listings, but actually understands what you want across trip planning, searching, and hosting.

Natural language search and Meta talent

A new natural language search feature is already quietly live for a tiny fraction of user traffic. Instead of clicking through dozens of amenity filters, travelers can now ask open-ended questions about listings and neighborhoods in plain English.

Chesky emphasized that this is just early testing. The plan is to make the entire search experience conversational and relevant across the whole trip. When analysts asked about sponsored listings inside AI search, Chesky said the priority is perfecting the user experience first, though conversational ad formats could come later.

To pull this off, Airbnb recently hired Ahmad Al-Dahle as CTO. Al-Dahle previously led Meta’s Llama model team, and he is tasked with putting Airbnb’s massive proprietary archive of verified identities and real guest reviews to work.

AI agents taking the wheel on support

Behind the scenes, Airbnb's AI is already doing serious heavy lifting. The LLM-powered support bot launched in North America last year now resolves a full third of customer issues completely on its own without routing to a human agent.

Chesky set an ambitious target for next year: significantly more than 30% resolution across all supported languages. Even bolder, support won't stay text-only for long; Airbnb plans to roll out voice support, letting guests actually talk to the bot on the phone.

Internally, 80% of Airbnb engineers already rely on AI tools daily, and leadership wants that number at 100%. The expansion comes alongside solid financials, with Q4 revenue rising 12% year-over-year to $2.78 billion.

Why it matters

Airbnb was smart to avoid the initial rush of gimmicky travel chatbots, and this calculated pivot feels much more grounded.

Generic chatbots are terrible travel planners because they don't know real-time inventory or nuanced property realities. Airbnb, on the other hand, sits on arguably the world's most valuable proprietary dataset of non-standard lodging reviews, house rules, and host interactions. Bringing in the former lead of Meta’s Llama team signals they intend to train specialized models directly on that data moat.

For travelers, conversational search could eliminate the tedious chore of hunting through hundreds of reviews just to confirm if a driveway can fit an SUV or if the stairs are safe for toddlers. But high stakes bring real risks: hallucinations around amenities or cancellation policies will quickly burn user trust, and introducing conversational ads down the line could make unbiased recommendations tricky to preserve.