Redfin operates a machine learning recommendation engine that personalizes the home listings each user sees — on the website, in the app, and in emails — based on their past browsing and engagement behavior. One of Redfin's earliest AI products, it now drives approximately 25% of all traffic to the Redfin website.

The engine analyzes signals including which homes a user clicked on, viewed in detail, favorited, or toured, as well as characteristics of homes that buyers with similar behavior ultimately purchased. The model outperforms users' own saved search filters at predicting which new listings they will engage with. Originally fully algorithmic, the system was updated to allow agents to manually add or remove homes from a client's recommendations, creating a hybrid approach where human judgment can override the model. Redfin's data advantage — access to actual tour and offer data from its brokerage operations — gives the model inputs unavailable to purely search-based platforms.