How Zillow’s new AI mode works throughout the real estate journey

Zillow's new personalized AI experience is designed to help buyers and renters ask questions in natural language, compare options and get help with what to do next. A coordinated AI system built for the complexity of home shopping powers that experience.

When a shopper asks a question, the system interprets intent, determines which capabilities are most relevant and composes a response grounded in real estate.

Rather than relying on a single AI model to answer every question directly, Zillow uses multi-agent architecture. A central agent coordinates a growing set of domain-specific "skills," including property search, financing and valuation, and invokes them as needed. Each skill combines data, logic and, in some cases, specialized models designed for a specific part of the housing journey.

This architecture allows the experience to go beyond general responses. A question about affordability, for example, can draw on financial context and product workflows, while a request to compare homes can combine listing data with broader signals about each property.

Routing intent across Zillow's ecosystem

Housing questions are rarely one-dimensional. Shoppers often ask layered questions that involve comparisons, trade-offs and timing.

To handle this, the AI system uses language models to interpret intent, then routes the request through the right combination of skills. This process, often described as tool or function calling, allows conversational input to connect to structured actions across Zillow's platform.

The result is a response that is not just generated, but assembled - shaped by both model reasoning and the output of Zillow's underlying systems.

Maintaining context across the journey

Home shopping unfolds over time. People revisit listings, refine preferences and build a clearer sense of what matters as they move closer to a decision.

To support this, the AI experience is designed to maintain context across both conversations and broader activity on Zillow. Rather than starting from scratch each time, it can draw on prior questions, browsing activity and recent interactions within a shopper's Zillow experience on the site and app to make responses more relevant and reduce repetition.

Technically, that means managing multiple forms of context, including the direction of the conversation and behavioral signals, so the system can adapt as a shopper's needs evolve.

Understanding homes beyond listing fields

Many of the details that shape housing decisions are not fully captured in structured data.

Zillow is addressing this by incorporating a deeper understanding of each home by using a combination of computer vision and multimodal modeling that converts photos and related data into a richer representation of that home. This allows the AI system to reason about attributes such as layout, light, finishes and condition, which are qualities that often influence shoppers when comparing properties.

By combining those signals with traditional listing data, Zillow can support more nuanced comparisons and recommendations, especially when shoppers are weighing trade-offs between homes.

Connecting conversation to action

A key distinction in Zillow's approach is how closely the AI experience is tied to the product itself.

Because it is integrated with Zillow's platform, the experience can connect conversation directly to next steps. Depending on the context, that may include viewing similar homes, exploring affordability, scheduling a tour or connecting with a real estate professional.

That integration is what allows the experience to move beyond answering questions and toward helping shoppers make progress in getting a home.

Designing for housing-specific safeguards

Housing is a regulated, high-stakes industry, and the AI experience is designed with that in mind.

Zillow incorporates safeguards that evaluate both user prompts and system responses, helping to ensure that outputs align with fair housing requirements and other real estate-specific constraints. These protections combine model-level prompting with internally developed classifiers and policies tailored to housing.

The goal is to provide helpful guidance while maintaining compliance and trust.

A coordinated system for a complex decision

Zillow has invested in AI across search, valuation, affordability and computer vision for years. This new experience brings these components together into a coordinated system designed to operate throughout the home-shopping journey.

Under the hood, the technology is not defined by any single model. It is the coordination of multiple systems - language models, skills, context management, home understanding and safeguards - working together in real time.

This is what allows Zillow's new AI experience to translate plain-language housing questions into structured guidance and actionable next steps.