Realtor.com
natural-language search with understanding of listing images
Realtor.com deployed in October 2025 a natural-language home search that recognizes more than 300 terms and analyzes listing photos, then launched RealAssist AI built with Google Gemini in June 2026.
Objective
Let the buyer phrase their search the way they speak, including complex criteria, and match their words to listing descriptions and photos to bring them faster toward a property and an agent.
The deployment
Realtor.com deploys in October 2025 a voice and text natural-language search that interprets complex queries, for example a single-story house with a pool in a given city and less than five years old. It also analyzes listing photos to surface matching properties, recognizes more than three hundred terms such as cathedral ceilings or a modern kitchen, and learns from the user's behavior to personalize the results. In June 2026, Realtor.com launches RealAssist AI, a conversational search experience built with Google Gemini and Google Cloud, available in beta to a group of logged-in users on desktop, app, and mobile web. RealAssist guides the buyer from their first budget questions through to being connected with a local agent, remembering preferences from one session to the next.
Results Proof C
Specialist press (HousingWire, Real Estate News) and an official Google Cloud release (T2) concordant on the deployment, with quantified feature characteristics. No isolated public engagement metric, hence a C level.
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 Google Gemini Models for the RealAssist AI conversational search (understanding the need, preference memory).
- infra Google Cloud Cloud platform on which RealAssist AI is built.
- outil Recherche NL in-house Realtor.com Voice and text natural-language search that recognizes more than 300 terms and learns from user behavior.
- outil Analyse d'images d'annonces Understanding of listing photos to surface properties matching the stated criteria; the exact vision model is not named.
How it runs, concretely
For ops teams-
1Spoken or written query customer
The user phrases their search the way they speak, including complex criteria.
-
2Text and image interpretation AI
The system recognizes more than three hundred terms and analyzes listing photos to find matching properties.
-
3Personalization and memory AI
It learns from behavior and, with RealAssist, retains preferences from one session to the next.
-
4Agent connection human
RealAssist guides from budget through to being connected with a local agent.
The richness of listings in text and images, plus user behavior. Without usable descriptions and photos, matching on precise terms degrades.
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
- listings with rich descriptions and photos
- reference of real-estate terms
- user behavior history
Org prerequisites
- AI team or cloud partner for the models
- computer vision capability on the photos
- personalization loop
Possible stack
- LLM for query understanding
- vision model for image analysis
- search and ranking engine
The plan, step by step
- Step 1Normalize listing descriptions and build the reference of real-estate terms for the target market.Deliverable: Structured vocabulary mapped to the listing inventory.
- Step 2Connect natural-language query understanding to the existing search engine.Deliverable: Internal beta of NL search on textual criteria.
- Step 3Index listing photos with a vision model to cover visual criteria (pool, single-story, modern kitchen).Deliverable: Image-to-criteria index connected to the search.
- Step 4Open a user beta and add personalization by learning from behavior.Deliverable: Public beta with search engagement measurement.
- Step 5Link the search to being connected with an agent and generalize.Deliverable: Complete query-to-property-to-agent journey, with a relevance dashboard.
First step: Enrich and normalize listing vocabulary before connecting image understanding.
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 2 more portals embrace AI-powered home search tools Secondary archive pending
- S2 Realtor.com launches AI-powered home search tool Secondary archive pending
- S3 Realtor.com Launches RealAssist AI: A Completely Reimagined Way to Find A Home Interested party 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.