Homes.com
natural-language voice and text search, then conversational assistant
Homes.com (CoStar Group) launched Smart Search in October 2025 then Homes AI in February 2026 for natural-language home search; the CEO says users of the AI mode spend more time on the site and submit more leads, on an audience of 108 million monthly unique visitors.
Objective
Replace filters with plain-language search, including multi-city and point-of-interest queries, to keep the user longer on Homes.com and increase the leads sent to subscribing agents.
The deployment
Homes.com, CoStar Group's residential portal, launched Smart Search on October 14, 2025: a natural-language search that accepts conversational queries such as a single-story house with a pool in a given city, multi-city searches, point-of-interest queries, and voice input through speech recognition. In February 2026, CoStar rolled out Homes AI, a search interface driven by a language model that enables two-way conversation by voice or text to explore properties and neighborhoods, with data that stays within the Homes.com ecosystem. The CEO reports that users who go through the AI mode stay longer on the site and submit more leads. Homes.com claims 108 million average monthly unique visitors in 2025.
Results Proof C
Statements from CoStar's CEO reported by trade press (Real Estate News) and consistent official CoStar releases. Qualitative engagement metric and a claimed audience, without a precise figure for the AI effect, hence a level 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
- plateforme Microsoft Azure OpenAI Cloud service that provides the conversational model behind the natural-language search and Homes AI; the exact model is not named publicly
- outil Smart Search / Homes AI (custom in-house) Layer built by CoStar: translation of sentences into search criteria, multi-city and point-of-interest queries, voice input (speech-to-text), then two-way conversation by voice or text
- infra Catalogue d'annonces et donnees de quartier CoStar CoStar Group's residential listings base and local data that feed the search; conversation data stays within the Homes.com ecosystem
How it runs, concretely
For ops teams-
1Voice or text query customer
The user describes their need in plain language, including multi-city or by point of interest.
-
2Interpretation and retrieval AI
Smart Search translates the sentence into criteria and surfaces the matching properties.
-
3Two-way conversation AI
In Homes AI mode, the user refines by voice or text to explore properties and neighborhoods.
-
4Lead generation human
Strong intent turns into a lead sent to a subscribing agent.
The listings catalog and the neighborhood data, plus the conversation data kept within the Homes.com ecosystem. Without them, the assistant can neither locate nor compare properties.
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
- structured listings catalog
- neighborhood data and points of interest
- controlled storage of conversation data
Org prerequisites
- access to a conversational model (cloud or internal)
- speech recognition
- routing of leads to agents
Possible stack
- conversational model such as Azure OpenAI
- NL search engine
- speech-to-text
The plan, step by step
- Step 1Structure the listings catalog and the neighborhood and point-of-interest data; define where the conversation data lives.Deliverable: Catalog queryable by criteria and a conversation-retention framework
- Step 2Build the natural-language search: translation of the sentence into criteria via a conversational model, testing on real queries including multi-city.Deliverable: NL search in production alongside classic filters
- Step 3Add voice input (speech-to-text) and measure time spent and search engagement against the filter journey.Deliverable: Voice mode active and a first engagement review
- Step 4Open up two-way conversation (property and neighborhood exploration) and connect lead routing to agents.Deliverable: Conversational assistant in production with lead-volume tracking
First step: Deliver a reliable NL search before opening up two-way conversation.
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 CoStar touts Homes.com momentum, bets big on AI after strong Q4 Secondary archive pending
- S2 2 more portals embrace AI-powered home search tools Secondary archive pending
- S3 Homes.com Launches Smart Search Feature, Enabling Users to Search the Way You Speak Primary 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.