Rightmove
natural language search plus genAI restyling of listing images
Rightmove rolled out AI Keywords in 2025, trained on 25 years of data, and Style with AI to restyle listing photos, then a conversational search on Google Gemini models in 2026; more than 80% of the time spent on UK property portals is spent on Rightmove.
Objective
Help the user find a property that matches their exact criteria in plain language, and help them picture themselves in a home to sustain engagement on the UK's leading property portal.
The deployment
Rightmove rolls out AI Keywords on its app in 2025: smart prompts such as exposed brick, river view, or underfloor heating as a visible feature, trained on twenty-five years of Rightmove data, that scan the images and text of listings to surface more relevant properties. The same year, Style with AI lets a buyer remove the furniture from a photo, adjust the lighting, and change a home's style, for example toward a Scandinavian or art deco look, to help them picture themselves in it. In February 2026, Rightmove opens a conversational search in beta through a Use AI button, where the user describes their need in natural language and refines it through conversation; this search is built with Google Cloud on Gemini models. Rightmove states that more than eighty percent of the time spent on property portals is spent on Rightmove and that it has twenty-seven AI initiatives in development.
Results Proof C
Official Rightmove releases (T1) and UK trade press aligned on the rollout, with evidence of scale (share of time spent, a 25-year database). No isolated effect metric, hence 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
- llm Google Gemini Models behind the Use AI conversational search (beta February 2026).
- infra Google Cloud Cloud platform on which the conversational search is built.
- outil AI Keywords (in-house Rightmove) Smart prompts trained on 25 years of Rightmove data that scan the images and text of listings.
- outil Style with AI (in-house Rightmove) GenAI restyling of listing photos (furniture removal, lighting, styles); the underlying generative model is not named.
How it runs, concretely
For ops teams-
1Search by prompt or conversation customer
The user picks a smart prompt or describes their need in natural language through Use AI.
-
2Scanning images and text AI
AI Keywords scans the images and text of listings to surface more relevant properties.
-
3Conversational refinement AI
Use AI search refines the results through conversation, on Gemini models.
-
4Visual projection AI
Style with AI removes furniture, adjusts lighting, or changes a photo's style to help buyers picture themselves.
The images and text of listings plus twenty-five years of Rightmove data. Without this corpus, the smart prompts cannot find the properties that carry the requested criterion.
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 usable images and text
- a long data history for training
- image rights for restyling
Org prerequisites
- an AI team or cloud partner for the models
- vision and image generation capability
- a high-traffic search surface
Possible stack
- NL search model
- vision model to scan listings
- generative image model for restyling
The plan, step by step
- Step 1Index the images and text of listings and build the reference set of searched criteria (exposed brick, underfloor heating, view).Deliverable: Multimodal listing index mapped to a vocabulary of criteria.
- Step 2Launch natural language prompt/keyword search on the app.Deliverable: NL search in production on a subset of criteria.
- Step 3Add the image restyling module (visual projection), after verifying photo rights.Deliverable: Restyling tool in beta with a validated legal framework.
- Step 4Build the conversational search on an LLM (refining the need through dialogue), with a cloud partner.Deliverable: Conversational beta with a user group.
- Step 5Measure relevance and engagement, then generalize.Deliverable: Relevance/engagement dashboard and roadmap of the next initiatives.
First step: Index listing images and text so that prompts find the right criterion before adding generation.
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 Rightmove unveils AI tools to enhance home search experience Primary archive pending
- S2 Rightmove upgrades AI tools to improve conversational property search Secondary archive pending
- S3 Rightmove launches next phase of AI-powered property search 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.