ATTOM launches AI model covering nearly 100 million US homes
ATTOM has expanded its Rental AVM rent estimation service with an AI-powered model that estimates current monthly market rents for nearly 100 million US homes. The company says the median absolute error is 5.9 percent for properties not yet listed for rent.
- The new Rental AVM estimates rents for nearly 100 million US properties.
- Coverage has expanded by about 40 percent.
- The median absolute error is 5.9 percent for homes not yet listed for rent.
- The model uses multiple machine learning models and cleaned rental listing data.
- The service is available through APIs, bulk data licensing, Snowflake, Databricks and the MCP Server.
Coverage expands significantly
ATTOM announced the Rental AVM expansion on November 3. The new model provides estimates for single-family homes, condos and townhouses in all 50 US states and the District of Columbia. Coverage has increased by about 40 percent, bringing the system to nearly 100 million properties.
According to the company, the model can also estimate rents for homes that have not yet been listed for rent. For these properties, the median absolute error is 5.9 percent. Alongside the estimate, the system provides a low and high estimated value, with the range narrowing as more evidence becomes available for the property.
The AI model draws on multiple sources
Rental AVM uses multiple machine learning models. They examine property characteristics, value, previous rental data and recent rental activity, then determine the weight of each factor based on the evidence available for the property.
The model was trained and validated using actual rental listings. Duplicate records, placeholder rents and implausible prices were removed from the dataset. Aaron Wagner, ATTOM's head of data science, said the system adapts to the rental information available for each property. According to the company, this enables it to produce estimates for homes with no rental history.
Rob Barber, ATTOM's chief executive, said: “Estimating rental values is difficult because lease agreements are private, and many homes have no rental history.” He said the development extends the approach behind ATTOM Intelligence, combining reliable property data, advanced AI and analytics, to more properties.
Investors, lenders and AI developers can use it
According to ATTOM, Rental AVM can support a range of business tasks. Real estate investors and landlords can use it to screen acquisitions, assess rental opportunities, compare rental income with property value and support rent-setting decisions.
Mortgage lenders and insurers can evaluate rental income for investment properties and DSCR loans, monitor portfolio exposure and support analysis of lost rental income. Real estate portals and AI developers can also integrate the estimates into listings, investment tools and AI-powered applications through ATTOM's MCP Server.
Rental AVM is updated monthly based on new rental activity and changing market conditions. The service is available through APIs, bulk data licensing, Snowflake, Databricks and the MCP Server. Based on the announcement, the expanded coverage could provide more usable data primarily to organizations that need to estimate rents for properties with no public rental history.
