An automated valuation model, or AVM, is software that estimates a home's value from data instead of a visit. Zillow's Zestimate is the best known. Lenders use AVMs every day, often without you knowing.
How they work
An AVM looks at recent sales, property records (size, age, lot), tax assessments, and sometimes listing photos. It finds patterns in what similar homes sold for and applies them to yours. Most use machine learning, which means they learn from millions of past sales.
How accurate are they?
It depends on whether your home is for sale. Zillow reports a nationwide median error of about 1.9% for homes on the market, because the list price gives the model a strong hint. For homes not for sale, the median error is about 7%. That means half of those estimates are off by more than 7%, which on a $400,000 home is more than $28,000.
Where they fall short
AVMs can't see inside your home. A new kitchen, a new roof, or water damage won't show up in public records. They also struggle where there are fewer recent sales or wider price swings.
Researchers at the Urban Institute found that AVM errors are larger in majority-Black neighborhoods. In their study published in HUD's Cityscape journal in 2024, the average error was 36.3% in majority-Black neighborhoods versus 13.8% in majority-white neighborhoods in their Atlanta data. Most of the gap came from differences in home prices and condition, but a gap remained even after the researchers accounted for those.
New rules for lenders
Since October 1, 2025, a federal rule requires mortgage lenders that use AVMs to have quality controls. The rule, issued by six agencies including the CFPB and FHFA, requires controls that:
- ensure a high level of confidence in the estimates,
- protect against manipulation of data,
- avoid conflicts of interest,
- require random sample testing and reviews, and
- comply with nondiscrimination laws.
That last one is new. It means lenders now have to check that their AVMs don't produce unfair results by neighborhood.
How Home Equalizer is different
We show our work. You see every sale we used, its match score, and how accurate values at that score have been in our testing. And because records can't see inside, your photos and upgrade list fill the gap.