Zillow's AI Strategy Hinges on Data Moat, Not Generic Models
Zillow is betting that proprietary data and transaction context will give it an edge over horizontal AI competitors in real estate. At an investor presentation, executives outlined how the company plans to build customized AI models trained on housing-specific information that generic large language models cannot match.
The strategy rests on a simple premise: Zillow sees the full arc of a real estate transaction. It tracks home searches, saved listings, neighborhood comparisons, mortgage rate checks, virtual tours, agent conversations, and closed deals. That visibility creates what executives call a "data flywheel" - the more users engage with Zillow, the more context the company gathers to train better AI.
CEO Jeremy Wacksman said 70% of people who buy or sell a home in the U.S. use Zillow during the transaction. Buyers spend two to three hours per week over five months researching. That behavioral data, combined with live housing inventory and pricing information, gives Zillow something competitors cannot easily replicate.
"Only Zillow can see things like all the homes you've saved and viewed and the ones you're coming back to every day signaling intent," Wacksman said.
Context as Competitive Advantage
During the three-hour presentation, Zillow executives mentioned "context" 24 times, "content" 39 times, and "data" 72 times. The repetition underscores where the company sees its advantage.
Zillow's AI mode - a new feature for consumers and agents - is designed to guide users through the buying or selling process by understanding their specific situation. A buyer who has viewed 50 homes, checked affordability calculators, and looked at neighborhoods multiple times receives different guidance than someone just starting their search.
Chief Technology Officer David Beitel said: "When we have the data and user intent we have and you think about the models we're building and training, who's going to build the best personalization? The ones who will build the best models to drive experiences are the ones that have the expertise and data to do that in context."
Zillow also owns rich media content - drone footage, 3D tours, floor plans - that feeds into AI training. The company recently launched SkyTour, which uses drone footage and AI to show properties and surrounding neighborhoods. These features generate data while improving user experience.
Real Estate Complexity as a Moat
Zillow executives argue that real estate's regulatory complexity creates a natural barrier against simpler competitors. The industry requires licensed brokerage relationships in all 50 states, with hundreds of thousands of brokers operating across over 500 multiple listing services and 1.5 million agents.
State regulations vary significantly. Some questions about property taxes, agency duties, or fiduciary obligations require licensed professionals to answer legally. Generic AI models cannot navigate this fragmentation.
Wacksman said: "The volume of complexity in this category is staggering. Every state is unique in regulation."
Zillow also operates backend systems for real estate agents through its Premier Agent service, which has 100,000 users. This means the company touches multiple points in a transaction - the consumer side and the professional side - creating a more complete picture than competitors.
The Challenge Ahead
Zillow faces serious competition. Rocket Companies, which owns Quicken Loans, recently acquired Redfin to build its own data flywheel combining mortgages, real estate services, and transaction data. That combination mirrors Zillow's integrated approach.
Nicholas Stevens, Vice President of Product, acknowledged the difficulty of getting AI right in real estate. "If you make up any of the answers, if you get anything wrong, if you give the wrong guidance, that's an incredibly trust-busting moment," he said. Generic models can surface listings, but Zillow's advantage lies in deeper guidance - understanding affordability, timing, and next steps.
The company has 300 people building AI-driven experiences. Internally, Zillow employees use coding tools like Cursor and Claude to accelerate development. The company is attempting to become what Co-Founder Lloyd Frink called "an AI native company."
Technology Stack and Partnerships
Zillow built its AI infrastructure on Amazon Web Services, using S3 for storage, Lambda for compute, SageMaker for model training, and a proprietary data lake called Zillow Data Lake. The architecture is designed to handle streaming data from millions of users and feed it into machine learning algorithms.
The company also partners with horizontal AI players like OpenAI and Anthropic despite viewing them as competitors. Frink said Zillow will meet customers wherever they are, whether on Google, OpenAI, or other platforms.
For real estate professionals, Zillow is layering AI into its product suite - Zillow Pro, Showcase, and Follow Up Boss - to automate content drafting, client messaging, document analysis, and lead scoring. Real-time tour customers convert at three times the normal rate, and preapproved buyers convert at five times the rate.
AI for Executives & Strategy professionals should note that Zillow's approach exemplifies how vertical integration and data analysis can create defensible competitive advantages in AI, even against larger horizontal competitors.
Your membership also unlocks:
