Airbnb is moving cautiously on flashy consumer AI, but it’s going all-in on AI as a build tool. According to TechCrunch AI, CEO Brian Chesky told investors on the company’s second-quarter earnings call that AI has cut the time from concept to launch by as much as 60% across key initiatives. That’s the headline most travel coverage buried, and it’s the one worth paying attention to.
What stands out here is the split. Airbnb is treating AI as an internal engine first and a customer feature second. Earlier this year the company said AI writes 60% of its code, TechCrunch AI reports. Now Chesky is putting a number on the payoff: nearly 80% more features and improvements shipped this year versus the same six months in 2025.
What’s actually changing
Chesky pointed to real areas where AI sped things up: search, sign-up, checkout, and payments. Plus host-facing tools like a quicker onboarding flow. This isn’t a demo. It’s velocity showing up in the roadmap.
The consumer side is deliberately slower. Airbnb has mostly stuck to review summaries and listing highlights, and Chesky has argued that a plain chatbot doesn’t fit travel. That’s a notable position while much of the industry races to bolt conversational interfaces onto everything.
Now the company is finally testing AI search. The key detail: it’s optional. Airbnb is adding a toggle so users can switch to natural-language search that returns visual results, without forcing anyone off the familiar filters.
“The titles [in the answer] could actually be AI-generated and they can be conversational as if you’re reading a chatbot, but more visual,” Chesky said. “Then you get to the product description page and the highlights are AI generated in real-time and personalized to you.”
The numbers behind the pitch
Where Airbnb has gone deep is support. The company launched an AI-powered bot in North America in 2025 and expanded it to more than 50 languages this year, with voice calls planned for later. The results, per TechCrunch AI:
- Nearly 45% of customer issues that start with the AI agent finish without any human help.
- Support cost per booking is down 16% year-over-year.
- Revenue rose 17% to $3.6 billion for the quarter ended in June.
- Adjusted EBITDA jumped 21% to $1.3 billion.
This is significant because it separates AI theater from AI economics. A 16% drop in support cost per booking is a hard operational number, not a keynote slide.
Why it matters now
A lot of companies are announcing consumer AI features to look current. Airbnb is doing something quieter and arguably smarter: using AI to compress its own build cycle and cut back-office cost, then rolling out customer features only where they clearly improve the experience.
There’s a caveat worth flagging. Self-reported figures like “60% of code” and “60% faster” come from the CEO on an earnings call, where the incentive is to sound ahead of the curve. Take the exact percentages as directional, not audited. The revenue and EBITDA numbers are the reliable anchors, and they’re strong.
The broader pattern is that AI’s clearest wins right now are internal. Faster shipping, cheaper support, less time from idea to launch. The consumer-facing magic is still catching up to the operational gains.
Takeaways for practitioners
If you’re deciding where to point AI in your own product or business, Airbnb’s playbook offers a few practical moves:
- Point AI at your build cycle first. Faster iteration compounds. It makes every future feature cheaper to test.
- Automate the expensive, repetitive back office. Support is the obvious target, and Airbnb’s cost drop shows the return.
- Give users an opt-in, not a takeover. The AI search toggle respects existing habits while testing the new interface. Low risk, real signal.
- Measure in dollars and hours, not demos. Cost per booking and time-to-launch are the metrics that survive scrutiny.
Airbnb’s next test, AI search behind a toggle, will show whether its go-slow approach to consumer AI pays off. For now, the real story is that AI is already reshaping how the company builds and supports its product. More details are available at the original TechCrunch AI report.