Medina County, Ohio, population 183,000, ran every residential property sketch through EagleView's AI last year. It cross-referenced aerial photographs against official assessment records for 145,467 parcels and flagged 70,198 discrepancies: decks that didn't appear on file, room additions that had never been assessed, garages that had quietly doubled in size sometime between the last two revaluations. County officials estimate the corrections added $35 to $40 million to the tax rolls and $12 to $14 million in new annual tax revenue, without hiring a single additional appraiser.
Nobody knocked on those doors. Nobody drove by with a clipboard. An algorithm scanned photographs taken from an airplane, compared what it saw against what the county had on record, and generated a discrepancy report that a handful of staffers processed at their desks.
That is the offense.
The defense looks like this: a startup called Ownwell raised $50 million in February 2026, bringing its total funding to roughly $74 million, to do the opposite. Ownwell's AI pulls comparable sales data, generates a ready-to-file appeal packet with evidence and argument, and submits it on the homeowner's behalf, charging nothing upfront. Ownwell takes 25 percent of whatever tax reduction it wins, and if it wins nothing, the homeowner pays nothing. Ownwell reports an 86 percent success rate across more than one million appeals processed, with an average annual savings of $774 per customer.
The Expected Value of Not Sitting This Out
Run the numbers on Ownwell's published statistics and the expected value of filing an appeal through their platform comes out to approximately $500 per year. That calculation: 86 percent success rate, multiplied by the $774 average savings, multiplied by the 75 percent the homeowner keeps after Ownwell's contingency fee. The product is $499.59. Round up.
Five hundred dollars. Zero downside risk. Zero upfront cost.
And yet only an estimated 5 to 10 percent of American homeowners appeal their property tax assessments in any given year. In Texas, where the appeal process is more accessible than most states, Ownwell processed appeals for over 200,000 properties whose owners had never protested at all, according to the company's disclosures. Two hundred thousand households leaving $500 on the table every year because nobody told them the expected value calculation, or because the process felt too opaque, or because they assumed the county was probably right.
On a $4,427 average annual tax bill, that $500 represents an 11.3 percent effective discount on the single largest recurring expense most homeowners face after their mortgage payment, their property insurance premium, and in some states their HOA assessment. Not guaranteed, obviously. But the expected value, across a large enough population of homeowners filing in good faith with reasonable comparable data, functions like a rebate that most homeowners never redeem because they do not know it exists.
What the County's AI Sees From the Sky
EagleView maintains a database of over three billion high-resolution aerial photographs covering 94 percent of the U.S. population. In February 2026, the company announced that its 3D property intelligence technology could measure walls, windows, and doors with 98.77 percent accuracy. That is not a typo. It resolves the dimensions of your home's exterior features from photographs taken at altitude with sub-two-percent error.
Riverside County, California, signed a five-year contract with C3 AI in 2023 to overhaul its assessment process entirely. Riverside consolidated from 30 separate appraisal models down to four unified AI models, cutting appraisal processing time by 40 percent while simultaneously applying those models to a county with 2.4 million residents and over 800,000 assessable parcels. New York City's Department of Finance piloted C3 AI for condominium revaluation, a politically sensitive undertaking in a city where assessment methodologies for condos and co-ops have faced criticism from both property owners and reform advocates for years. Cook County, Illinois, which includes Chicago, released PTAXSIM, a public simulator that lets residents model how assessment changes flow through to their tax bills. Wake County, North Carolina, deployed SAS Viya's machine learning platform to generate assessments using hundreds of property variables and daily sales data.
CAPE Analytics, acquired by Moody's, applies computer vision to aerial imagery to identify roof type, detached structures, vegetation overgrowth, and other physical characteristics that inform both insurance underwriting and assessment accuracy. Over 80 enterprise clients use the platform, and when Moody's bought the company, it merged property-level AI with catastrophe risk models, creating an assessment-and-risk intelligence layer that individual counties could never build on their own.
The Equity Problem Nobody Wants to Quantify
An algorithmic arms race in property taxation sounds like an efficiency story until you look at who wins and who loses when one side gets better tools faster than the other.
Research from the Minneapolis Federal Reserve found that Black and Hispanic homeowners face assessment rates 10 to 13 percent higher than white homeowners within the same jurisdiction, translating to $300 to $390 per year in excess tax burden. Separately, the Brookings Institution estimated that homes in majority-Black neighborhoods are undervalued by 21 to 23 percent relative to comparable homes in non-Black neighborhoods, representing $162 billion in aggregate lost property value across 113 metropolitan areas.
Those two findings point in different directions, they appear to contradict each other, and yet they are both correct simultaneously. Homes in predominantly Black neighborhoods carry depressed market values (Brookings) while simultaneously being over-assessed relative to those depressed values (Minneapolis Fed). Owners pay more in tax as a percentage of what their home is actually worth, bearing a disproportionate burden that compounds every year the assessment ratio remains uncorrected.
Minority homeowners are less likely to appeal their assessments, and when they do appeal, they win less often and receive less financial relief than white homeowners filing comparable claims in the same jurisdictions. Northwestern University field experiments showed that providing step-by-step guides and information about the appeals process increased minority appeal rates, but the gap in outcomes persisted even after accounting for participation differences.
An AI appeal tool that charges nothing upfront and files automatically could, in theory, close that gap. Or it could widen it, if adoption patterns mirror every other financial technology product released in the last decade: early adopters skew wealthier, whiter, and more digitally connected, and the population that would benefit most learns about the tool last.
The Strongest Case for the County's Side
Counties are not villains in this story, even if the aerial surveillance framing makes them easy to cast as one. Accurate assessments fund schools, fire departments, road maintenance, and the basic municipal infrastructure that keeps a neighborhood functioning. When a homeowner builds a $40,000 addition and never pulls a permit, every other taxpayer in the jurisdiction absorbs the cost of the services that addition consumes. Neighbors whose assessments accurately reflect their improvements end up subsidizing the neighbor whose deck materialized overnight without a permit number attached to it. The county's AI corrects a free-rider problem that manual inspection never could at scale.
EagleView's defenders argue, with some justification rooted in actual deployment data, that algorithmic assessment reduces the human bias that has plagued property taxation for a century. A human assessor driving through neighborhoods makes subjective judgments influenced by curb appeal, landscaping quality, and street condition, all of which correlate with the racial and economic demographics of the block being assessed. A satellite does not care who lives in the house, does not register curb appeal, and measures the addition at 200 square feet regardless of whether the homeowner is a partner at a law firm or a retired schoolteacher who speaks limited English and would never think to challenge a tax bill.
And then there is the collective-action problem that nobody on the startup side wants to discuss: if algorithmic appeals become universal, and every homeowner uses an AI service to minimize their assessed value, the tax base contracts. The same amount of revenue must still be collected from a reduced assessment base, which means the mill rate increases, every homeowner's effective rate climbs, and the primary beneficiaries of the entire exercise turn out to be the appeal-service companies that captured 25 percent of the savings on the way through. Counties in Texas have already begun hiring their own AI defense against appeal algorithms, and at least one competitor, TTP, filed suit against Ownwell in the Northern District of Texas in June 2025, alleging that Ownwell's quoted 2.5 percent tax rate differed from its algorithmic rate of 2.25 percent.
What This Means If You Are Building or Buying
If you are buying new construction, your home will be assessed for the first time using whatever AI tools your county has deployed. In Riverside County, that means C3 AI's unified models. In jurisdictions using EagleView, it means your home will be photographed from the air and measured to sub-two-percent dimensional accuracy before you receive your first tax bill. The assessment will not be a rough estimate derived from neighborhood comparables and a site visit that lasted four minutes. It will be a precise, algorithmically generated valuation that incorporates every visible feature of your property and cross-references it against every recorded sale in the jurisdiction.
If you are building an addition or making improvements, assume the county's AI will detect the change within one to two assessment cycles, because the days of finishing a basement, adding a bathroom, or enclosing a porch and hoping nobody notices are functionally over in any county that has contracted with EagleView, CAPE Analytics, or a similar aerial intelligence provider. You should pull the permit regardless. But the practical consequence of not pulling it has shifted from "maybe someone drives by and catches it" to "an algorithm will flag the discrepancy in the next flyover, and the resulting correction will include back-assessed taxes plus penalties."
If your current assessment looks high, file an appeal using one of the AI services if you want it done for you, or pull your own comparable sales data from your county's records and file directly. In contingency-fee jurisdictions, the expected value math strongly favors appealing: $500 in expected annual savings against zero downside, compounding every year you own the home.
Limitations
This analysis relies on Ownwell's self-reported success rates and savings figures. No independent audit of those numbers exists, and the company's incentive to report favorable statistics is obvious. Ownwell's $774 average savings figure is a national average that masks significant regional variation. In jurisdictions with low effective tax rates, the savings may not justify even a zero-cost appeal process when measured in time and attention. EagleView's Medina County case study is a single-county result in a state with specific assessment practices, and the $35 to $40 million figure represents gross additions to the tax rolls before successful appeals reduced the net impact. We could not determine how many of those 70,198 flagged discrepancies were contested or overturned. The racial disparity data from the Minneapolis Fed and Brookings uses methodologies that may not fully control for property condition differences within assessment districts.