A Computer Never Met Your Neighborhood. It Might Still Be Underpricing It.
Automated valuation models now price millions of home sales, refinances, and equity lines without a human ever walking through the door. Research shows they can still land on the same lower numbers traditional appraisals did.
Automated valuation models, or AVMs, were supposed to be the fix. Remove the human appraiser, the thinking went, and you remove human bias along with the subjective judgment calls that research has repeatedly linked to lower valuations in Black and Latino neighborhoods. It’s a reasonable idea. It’s also, according to recent research, an incomplete one.
Find a Mortgage Professional →
What is an automated valuation model, and how does it actually work?
The appeal is obvious: an AVM can produce a number in seconds instead of the days a full in-person appraisal takes, and it costs the lender far less. The trade-off is that a computer model can’t walk through your renovated kitchen, notice a foundation crack, or account for details a human would catch, and it can only be as fair as the historical sales data it was trained on.
How much lower do AI and traditional appraisals value homes in Black neighborhoods?
This is the uncomfortable core of the issue: AVMs are trained on historical sales data, and that data already reflects decades of the very bias they were meant to correct. A model that learns “homes in this type of neighborhood typically sell for X” will keep predicting X, even when X was shaped by discriminatory practices in the first place. Freddie Mac’s own research has documented that the gap between appraised value and eventual sale price widens as the share of Black or Latino residents in a neighborhood increases.
Can I dispute an AI-generated home valuation?
Federal regulators, including the Consumer Financial Protection Bureau, have acknowledged that both human and algorithmic appraisals carry fair-lending risk, and have pushed for stronger quality-control standards on AVMs specifically. That regulatory attention gives homeowners real, legitimate ground to push back on a number that doesn’t reflect their home’s true value or condition.
What should I do before my next appraisal, refinance, or sale?
| Step | Why It Matters |
|---|---|
| Ask your lender if an AVM or a full appraisal will be used | Some loan types allow you to request a full in-person appraisal instead |
| Ask for the AVM’s confidence score | A low confidence score is a strong signal to request a fuller review |
| Document recent upgrades and comparable sales yourself | Neither AVMs nor rushed in-person appraisals always catch every improvement |
| Work with a realtor or lender who understands this issue | An advocate who knows the appeal process can meaningfully change the outcome |
✅ Free Tool: Appraisal Dispute Readiness Checklist
Check off what you have in place before your next valuation.
Frequently Asked Questions
What is an automated valuation model (AVM)?
An AVM is a computer-driven formula that estimates a home’s value using property records, recent local sales, and market data, without a human appraiser inspecting the property in person.
Do AVMs actually reduce racial bias in home valuations?
Not fully. January 2026 Urban Institute research found AVMs produce errors 3.4 percentage points higher for Black homeowners than white homeowners, since the models are trained on historical sales data that already reflects existing bias.
Can I request a full in-person appraisal instead of an AVM?
In many cases yes, depending on your loan type and lender, and this is a reasonable request if you’re concerned about an AVM’s accuracy for your property.
How do I dispute a home valuation I believe is too low?
Provide your lender with documentation of recent improvements and comparable recent sales, and formally request a reconsideration of value through your lender’s appeal process.
Browse the Directory →