Zillow
natural-language search over a listings catalog
Zillow launched natural-language home search in January 2023, presented as a first for a major residential portal, enriched in September 2024 with commute time, budget, schools, and points of interest.
Objective
Let the buyer or renter describe their ideal home in one sentence rather than juggle filters, to surface relevant listings faster and shorten the search.
The deployment
Zillow launched natural-language search on January 26, 2023 on its iOS app, presented as a first for a major residential portal. The user types a sentence such as a price and neighborhood goal with a specific criterion, and the system scans millions of listing details to surface relevant properties. On September 4, 2024, an enriched version added search by commute time, monthly budget, schools, and points of interest, with queries such as a maximum commute time from a location or a home near a train station. Machine learning models parse the query, personalize the results, and learn to respond better to human sentences. The user can save a search and be notified when a new listing matches. Zillow is described as the most-visited real estate site in the United States.
Results Proof C
Official Zillow releases (T1) and specialized press agree on a rollout at the scale of the leading US portal. Without an isolated, public engagement metric for this specific feature, the level stays C.
How it works
Inferred typical approachThe internal detail is not public. Here is a proven approach that leads to the same result, to adapt to your stack.
The stack in detail
- llm Modeles NLP de comprehension de requete (in-house Zillow) Traduisent la phrase de l'utilisateur en criteres de recherche (prix, quartier, trajet, ecoles) ; l'architecture exacte n'est pas publiee.
- outil Moteur de recherche et de classement Zillow Balaie des millions de details d'annonces et personnalise l'ordre des resultats.
- infra Catalogue d'annonces et donnees locales Annonces, temps de trajet, ecoles et points d'interet ; leur fraicheur conditionne la pertinence des resultats.
How it runs, concretely
For ops teams-
1Natural-language query client
The user describes their ideal home in one sentence in the search bar.
-
2Interpretation and scan AI
The ML models translate the sentence into criteria and scan millions of listings.
-
3Result personalization AI
The system ranks the most relevant properties by the expressed preferences.
-
4Save and alert client
The user saves the search and gets a notification when a matching listing appears.
The quality and freshness of listing details and local data (commute, schools, points of interest). Without them, the user's sentence does not translate into relevant filters.
How your customers perceive this type of use
Sourced studiesLes consommateurs n'acceptent pas les chatbots par defaut : 64% prefereraient que les entreprises n'utilisent pas d'IA dans leur service client (Gartner, 2024) et pres d'un utilisateur sur cinq du service client par IA n'en retire aucun benefice (Qualtrics, 2025). L'acceptation se construit sur trois conditions mesurees par Salesforce : savoir qu'on parle a une IA, pouvoir escalader vers un humain, comprendre la logique de l'agent.
Acceptance conditions
- Etre informe qu'on parle a une IA et non a un humain (pres de 75% le demandent, Salesforce 2024)
- Un chemin d'escalade clair vers un agent humain (45% plus enclins a utiliser l'agent IA, Salesforce 2024)
- Une logique de l'agent clairement expliquee (44% plus enclins, Salesforce 2024)
Red lines
- Rendre l'humain injoignable : c'est la premiere inquietude des consommateurs sur l'IA dans le service client (Gartner 2024) et 50% craignent que l'IA les coupe du contact humain (Qualtrics 2025)
- Remplacer le service client par l'IA sans alternative : 53% envisageraient de partir chez un concurrent (Gartner 2024)
Sources: Salesforce 2024 · Gartner 2024 · Qualtrics 2025
How to replicate
Inference, not sourcedData prerequisites
- rich, structured listings catalog
- local geographic data (commute, schools, points of interest)
- behavioral history for personalization
Org prerequisites
- NLP and search team
- large-scale search infrastructure
- saved-search and notification system
Possible stack
- NL query understanding model
- search and ranking engine
- notification pipeline
The plan, step by step
- Step 1Structure the catalog and local data (commute, schools, points of interest) into queryable attributes.Deliverable: Enriched catalog with normalized attributes.
- Step 2Build the query understanding model (sentence to criteria).Deliverable: Parsing evaluated on a corpus of real queries.
- Step 3Connect the parsing to the search engine and ranking.Deliverable: Relevant results validated in an internal beta.
- Step 4Open the public beta on one platform (mobile app first).Deliverable: Engagement and saved-property measures.
- Step 5Add saved searches and notifications, extend the criteria covered.Deliverable: Active retention loop (new-listing alerts).
First step: Translate a free-form sentence into reliable search criteria before adding personalization.
The same pattern, proven in other industries
Amazon
genAI shopping assistant
Booking.com
genAI trip planning assistant
Burberry
genAI copilot for consultants (search, image recognition, complete-the-look)
Sources
- S1 Zillow's new AI-powered natural-language search is a first in real estate Primary archive pending
- S2 Zillow's AI-powered home search gets smarter with new natural language features Primary archive pending
- S3 Zillow, Unfiltered: Portal Giant Adds AI-Fueled Natural Language Feature Secondary archive pending
An error, newer info, a source?
This page lives on its accuracy. If a figure has moved, if the deployment has changed, or if you have a higher-quality source, tell us. Every sourced correction is verified before publication.