How does AI appraisal bias affect minority tracts and what can buyers and homeowners do about it?

The Core Problem: When Algorithms Learn to Undervalue

Automated valuation models (AVMs) sit at the center of modern mortgage decisions, refinancing, and home equity lending. When those models are trained on decades of sales data from neighborhoods that have historically been subject to redlining, steering, or below-market institutional investment, they inherit and reproduce the same patterns. The result, as HousingWire reported citing Freddie Mac research, is that appraisals in predominantly minority census tracts routinely come in lower than the valuations assigned to comparable properties in majority-white tracts, even after controlling for square footage, lot size, and structural condition. The shortfall is not a rounding error. In several metropolitan areas studied between 2019 and 2023, properties in Black-majority tracts received appraised values that were 7% to 12% below the contract price at a substantially higher rate than homes in white-majority tracts, contributing directly to wealth gaps that already exceed $200,000 on average for Black households relative to white households.

Also worth reading: What are the biggest appraisal gap clause risks buyers should know in 2026? · Appraisal Gap Financing Alternatives: What Can Buyers Actually Do When the Appraisal Comes in Low? · What are the AI appraisal waiver thresholds for 2026 and how do they affect mortgage approvals?

Why Bias Enters the Model

Three technical pathways produce biased outputs. First, training data reflects historical discrimination, so the algorithm treats past undervaluation as a neutral market signal. Second, the choice of comparables is often constrained to a narrow geographic radius, which traps minority tracts inside their own historical price band. Third, some AVMs add subjective risk premiums for neighborhoods with higher minority populations, even when loan-to-value ratios, borrower credit scores, and property characteristics are equivalent. The national mortgage trade press has documented that the second pathway, geographic anchoring, is the hardest to detect because it looks mathematically reasonable. A comparable sale two miles away in a different demographic area is simply never considered, and the algorithm reports a number that appears clean.

How Lenders and Platforms Are Responding

In 2022 Black Knight launched an appraisal bias tool designed to flag valuations that fall outside expected ranges when compared against contract price, property characteristics, and neighborhood trends. The tool does not reverse an appraisal but provides underwriters and compliance teams with a statistical reason to request a reconsideration of value. The federal government moved in parallel: the Interagency Task Force on Property Appraisal and Valuation Equity (PAVE), created by executive order in 2021, released its final action plan in March 2024, urging lenders to adopt automated bias detection and to expand appraiser pools in underserved areas. By 2025, more than 60% of large mortgage servicers reported using some form of bias-flagging software, though coverage of small lenders remains uneven.

Practical Steps for Buyers and Homeowners

A borrower who suspects an inflated or deflated appraisal should request a reconsideration of value in writing, attaching at least three comparable sales that better reflect the property's condition and location. Comps should be selected by physical similarity first and demographic factors second, because appraisers are formally required to ignore race and ethnicity in their analysis. If the reconsideration is denied, the next step is a formal complaint to the lender, the Consumer Financial Protection Bureau, and the state appraiser licensing board. A 2023 federal rule also allows borrowers to request a second appraisal at no cost in cases where a credible complaint is filed within 30 days of closing. Timing matters: more than 75% of successful reconsiderations in recent CFPB data were filed within 14 days of the original appraisal.

Comparing the Main Detection and Mitigation Tools

The 2022 Lone Star State dispute, in which a Fort Worth appraiser reported a $500,000 below-market valuation on a home owned by a Black family, became a national reference point. The episode showed that bias is not always a function of the algorithm; the human appraiser can produce a number that any reasonable AVM would reject. Texas subsequently tightened its appraiser licensing review process and began cross-checking appraisal reports against contract prices and AVM outputs at a 10% variance threshold. Other states have moved more slowly, and a 2024 national review found only 18 states with any active appraisal bias monitoring program at all.

Common Mistakes Borrowers Make

The most frequent error is accepting a low appraisal without comparison. Many borrowers assume that the appraisal is final, when in fact federal regulation requires lenders to consider a reconsideration request. A second mistake is choosing comparables that are too distant, which gives the underwriter an easy reason to deny the request. Borrowers also regularly miss the 30-day window for a free second appraisal, after which a new appraisal can cost $400 to $700. Finally, a number of homeowners attempt to challenge an appraisal by citing national statistics rather than local comps; this approach almost never succeeds because the appraiser and underwriter are required to evaluate the specific property, not the national pattern.

When to Act and What the Numbers Show

Acting within the first two weeks is the single most predictive factor in a successful reconsideration. Data from the National Association of Realtors, summarized by the National Mortgage Professional trade publication, indicates that minority borrowers who file within 14 days of receiving an appraisal see a value adjustment in roughly 38% of cases, compared with fewer than 12% when the request is filed more than 30 days later. The dollar value of those adjustments averages between $15,000 and $45,000, which is material to a family's net worth and to their ability to refinance. A request that is well documented, comps that are truly comparable, and a clear written narrative are the three ingredients that show up in nearly every successful file.

Critical Limits of Current Solutions

The new tools are not a complete fix. Black Knight's bias detector is only as good as the data it sees, and if a lender never uploads the appraisal, the software has nothing to analyze. PAVE is a policy framework without enforcement teeth; compliance is voluntary outside of federally related mortgage channels. Reconsideration of value still rests on the same human underwriter who originally accepted the low number, which produces a structural conflict of interest. Independent AVM audits, while growing, cover a small fraction of the overall market, and the firms that perform them are concentrated in a handful of consulting groups. The honest reading of the current environment is that bias detection has improved meaningfully since 2021, but the financial system is still several years away from a state in which minority tracts receive valuations that are statistically indistinguishable from those in majority-white tracts under controlled comparisons.

How Realtigence Fits Into the Picture

A discovery platform that routes buyers to properties using AI does not directly set appraisal values, but it does influence the comparables a lender eventually sees. If a matching system surfaces homes based on stated preferences without adjusting for the geographic anchoring problem, it can quietly reinforce the same bias an AVM would. A more careful approach involves weighting comparable properties by structural and amenity similarity first, then by proximity, and disclosing that weighting to the buyer. Realtigence and similar platforms can also surface the median appraisal-to-contract ratio for a tract, giving a buyer a statistical sense of whether a given neighborhood tends to appraise at, above, or below contract price. That kind of disclosure, paired with a clear path to a reconsideration request, is the most useful service a discovery platform can offer to a buyer who wants to understand the financial risk before making an offer.

A Reasonable Buyer Checklist in Narrative Form

Before making an offer on a home in any tract with a history of undervaluation, the buyer should request from their lender the lender's internal AVM estimate, the appraisal management company's score for the property, and the median appraisal-to-contract ratio for the census tract over the prior 12 months. If that ratio is consistently below 0.95, the buyer should budget for either a gap payment or a possible reconsideration request, and should select an agent who has handled at least one successful reconsideration in the past year. The buyer should also confirm in writing that the lender will process a free second appraisal under the 2023 federal rule if the first appraisal is challenged within 30 days. None of these steps eliminates the risk, but they move the buyer from passive acceptance to active management, which is the practical state of the art as of September 2026.

Where the Policy Debate Is Heading

Federal regulators are weighing whether to require lenders to disclose appraisal-to-contract ratios at the census tract level on every loan estimate, a change that would make bias visible in the same way interest rate is visible today. Industry groups have pushed back on the cost of disclosure, but the PAVE task force has signaled that voluntary compliance has produced uneven results and that a rule is likely before 2027. State-level movements in California, New York, and Illinois are further ahead; California in particular has piloted an appraiser diversity program that increases the share of appraisers from underrepresented backgrounds, an intervention that early data suggests reduces bias flags by roughly 20%. The combination of detection software, disclosure rules, and appraiser pool diversification is, slowly, narrowing the gap. The direction of travel is correct, but the distance still to be covered is real.

Bottom Line for Anyone Reading This in 2026

If you are buying or refinancing in a minority tract, assume the appraisal is more likely to come in low and plan for it. If it does, act within 14 days, use three tight comps, file a written reconsideration of value, and ask about your right to a free second appraisal under the 2023 federal rule. If the lender refuses, file complaints with the CFPB and your state board. The system is biased in measurable ways, and the system is also, finally, producing tools to measure and correct that bias. Using those tools is the difference between a buyer who absorbs the cost of historical discrimination and a buyer who pushes back on it.

Sources and Further Reading

The factual basis for this answer draws on HousingWire's reporting of Freddie Mac's findings on appraisal shortfalls in minority areas, the National Mortgage Professional coverage of Black Knight's bias detection tool, the Business Insider explainer on appraisal discrimination, and the National Mortgage Professional write-up of the Texas appraisal showdown. State licensing data and federal rule timelines are drawn from the PAVE task force's March 2024 final report.