Airbnb to roll out LLM features for search, trip planning

Airbnb plans to integrate large language model–powered features across its platform, according to CEO Brian Chesky during the company’s Q4 earnings call.

The company aims to enhance search, trip planning, and host management tools, with current testing of a natural language property search feature.

Chesky indicated that AI-driven customer support, which now handles about a third of issues in North America, will expand to include voice support and more languages.

🔗 Source: TechCrunch

🧠 Food for thought

Implications, context, and why it matters.

Airbnb’s AI support followed a deliberate customer support first strategy

While rivals shipped AI trip planners, CEO Brian Chesky started with customer support. He called it the “hardest problem” given the high stakes and low tolerance for hallucinations or error 1 2.
This push started years ago with an internal “Intelligent Automation Platform,” built to power conversational AI tools across channels such as in-app messaging and automated phone calls 3.
Airbnb engineers found that large AI models can run too slowly and cost too much for real-time support, which made response time (latency) a tight design limit 4.
In an A/B test, a smaller fine-tuned model cut average manual processing time by 13%. A larger model using chain-of-thought (a prompting method that encourages step-by-step reasoning) raised it by 3%, which the paper linked to higher latency 4.

For Airbnb, AI works as a defense against disintermediation

Chesky sees generic AI chatbots like ChatGPT as top-of-funnel sources. He told investors that traffic from chatbot platforms “converts at a higher rate” than traffic from Google 5.
To avoid becoming a simple booking link inside an AI-made itinerary, Chesky said Airbnb is weaving AI through the experience in a way he argued would be “impossible to replicate.” He called it the company’s “strongest defense against disintermediation” (when another platform inserts itself between Airbnb and customers) 5.
Chesky cited Airbnb’s proprietary advantages, including its verified identity base, and argued a generic chatbot cannot copy what Airbnb provides. Separate analysis credits proprietary datasets, for example listings, reviews, and availability, as part of travel platforms’ moats 2 6.
Chesky tied this specialized AI to a larger goal of turning Airbnb into an “everything” platform for travel and living. He compared the ambition to Amazon’s expansion beyond books 7.