I led the ideation and design of a first to market blended AI search experience, shaping how we balanced innovation with the realities of search, scale, and conversion.
The Opportunity
When guests search for vacation rentals, they don't think in filters and checkboxes. They think in stories like "somewhere cozy for a family reunion" or "a place with character near great restaurants." Traditional search can't handle that complexity.
So we built something that could.
TL;DR
First to market blended AI search experience
350,000+ AI searches in first 6 months
10% step conversion lift
4% increase in saved homes
54 million media coverage impressions
Blended AI experience featuring AI natural language search, AI suggested filters and AI generated destination content
The Approach
We assembled a small cross functional team of product, design, and engineering leaders with a clear mandate. Be first to market with an AI experience without harming the core business.
Our objectives were deliberate. Boost brand visibility, prove technical and financial feasibility, gather behavioral data and most critically, protect the existing booking funnel. Every decision would be measured against conversion rates and revenue impact.
With those goals in mind, we got to work. But we didn't get it right on the first try.
Doing Immediate Harm
Like many teams exploring AI in 2023, we started with what seemed obvious, an AI powered chat experience. We believed giving users conversational AI to plan trips and discover destinations was the future.
The reality? Every round of user feedback revealed fundamental issues.
Chat bots carried the stigma of customer service frustration. Users associated them with desperate attempts to cancel subscriptions or repeated pleas to speak with a human. Chat added unexpected steps between homepage and search results. People who wanted to simply search for a destination found themselves in an unwanted conversation.
No matter how seamlessly we tried to embed it, chat complicated the otherwise straightforward process of searching for a destination, specifying dates, and hitting search.
The Pivot Point
While the team explored chat based ideas, I made a deceptively simple observation. The difference between chat and search was only about 800 pixels.
Moving the text input field from the bottom of the page, which signals chat, to the top, which signals search fundamentally changed the user's mental model. It was the breakthrough we'd been looking for. Instead of injecting AI chat into the funnel, we could enhance the search experience with AI.
That unlocked a larger opportunity. We used the insight to redesign search from the ground up, rethink information architecture, reimagine how users discover properties, and create a foundation for AI capabilities beyond natural language input.
Reimagining search from the ground up allowed us to elevate and modernize the experience while building a proper foundation for the AI implementation.
The Design Solution
We nearly made the same mistake auto manufacturers have been making for years, combining everything into one "smart" interface that actually complicates the experience.
We could have built a single text field that handled both traditional destination searches ("Paris") and natural language queries ("romantic getaway with Michelin star restaurants"). Technically feasible? Absolutely. Good for our users? Maybe not.
The insight was simple. Users expect different tools for different tasks. Sometimes they want traditional search. Sometimes they want AI discovery.
The solution was a tabbed interface allowing users to seamlessly switch between Destination Search (familiar, fast, direct) and AI Search (exploratory, descriptive, discovery-focused.)
This choice also had business implications beyond UX, as it allowed us to strategically manage costly AI API calls by creating natural drop off points while giving users clear control.
By building the listing cards and destination info into a persistent drawer we were able to create an experience equally weighted for guests who prefer interacting with a map and guests that prefer scrolling through listings.
Keeping An Eye On Adoption
We created a journey of discovery.
We rotated prompt examples in the search field like "Rustic cabin by serene lake, close to fishing dock, kayaks, hiking trails" and "Romantic getaway with quaint local restaurants in Europe." We introduced seamless guidance that broke down important prompt elements like location, trip type, activities and home features. And we transformed loading screens into learning opportunities with "Trending Searches" and "Search Tips."
Moving beyond static filters, we implemented contextual, AI-suggested filters that understood search intent.
This wasn't just a convenience feature, it demonstrated that AI understood guest intent and provided us with valuable data insights. 35% of users selected "Waterfront" when suggested. 31% engaged with "Private Pool." 19% clicked "Pets Allowed."
Our most innovative departure from conventional search was the introduction of AI-generated location recommendations with rich, contextual information.
For guests interested in exploring possibilities, we provided multiple destination suggestions based on their search along with detailed location information across 5 categories, Culture, Kids, Nature, Weather and Travel.
The strategic impact? Previously, users left our site to research destinations on Google, potentially landing on competitor sites. We internalized this entire discovery journey.
Discovery through progressive education. Search examples, loading state tips, and AI generated location information.
The Results
Conversion
10% increase in AI Search to PDP conversion compared to traditional search
4% increase in AI Search to Saved Homes conversion compared to traditional search
7% increase in overall Search to PDP step conversion
1.75% increase in overall Search to Saved Homes conversion
Engagement & Adoption
170,000+ AI searches in first 100 days (March-June 2024)
350,000+ total AI searches through August 2024
10 words average prompt length (range: 1-188 words)
11% YoY increase in search page visits post-launch
14% YoY increase in overall site visits post-launch
10% increase in AI Search to PDP conversion compared to traditional search
4% increase in AI Search to Saved Homes conversion compared to traditional search
7% increase in overall Search to PDP step conversion
1.75% increase in overall Search to Saved Homes conversion
Engagement & Adoption
170,000+ AI searches in first 100 days (March-June 2024)
350,000+ total AI searches through August 2024
10 words average prompt length (range: 1-188 words)
11% YoY increase in search page visits post-launch
14% YoY increase in overall site visits post-launch
Brand Impact
First to market blended AI search experience in the Travel & Hospitality industry
54+ million impressions across media coverage
48 pieces of press coverage
54+ million impressions across media coverage
48 pieces of press coverage
The desktop experience featuring natural language AI search, AI suggested destinations, AI generated location insights and predictive AI filters.
Reflections
We set out to be first to market. And we were. But the more important discovery was that AI in product design works best when it enhances familiar patterns rather than replacing them.
The most successful AI features feel like natural evolution, not revolution. And sometimes the best AI experience is the one users don't even realize is powered by AI at all. They just say, "Finally, something that works the way I think."