Right now, a computer vision algorithm is staring at your roof from orbit. It has cataloged the age of your shingles, the square footage of moss growing along the north slope, the overhang of the oak tree your kids climb, and the patch job your roofer swore would last another five years. It has assigned your property a numerical score that your insurer has already seen, factored into its renewal decision, and acted upon, all before you received any notification that the assessment existed.
Welcome to the age of AI-scored homeowner insurance, where a satellite image taken on a Tuesday can trigger a non-renewal letter that arrives on a Friday, and in most states there is no law requiring the insurer to show you the image, give you time to fix the problem, or explain which algorithm made the call.
How the Scoring Works
Cape Analytics, founded in 2014 and acquired by Moody's in 2024, built the category. Computer vision models trained on geospatial imagery analyze every insurable property in the United States, extracting a targeted set of attributes: roof condition, roof covering material, roof geometry, tree overhang, and total structure area. Each attribute feeds a loss-predictive signal that insurers incorporate directly into underwriting and pricing decisions.
Cape's own analysis, validated across millions of aggregate carrier-contributed loss years, found that roofs scored severe or poor experience 2.5 times the wind and hail claims frequency of roofs scored excellent. Homes with large amounts of foliage overhanging the roof experience 90 percent higher wind-related losses. Those are not marginal differences, and they are the kind of numbers that make an underwriter's pen stop mid-signature on a renewal that would have been automatic three years ago.
As of early 2026, Cape's geospatial property attributes have been included in Department of Insurance-approved rate filings across 15 states: Arizona, Colorado, Illinois, Indiana, Iowa, Kentucky, Michigan, Mississippi, North Carolina, Ohio, Pennsylvania, South Dakota, Tennessee, Virginia, and Wisconsin. Hartford, Hippo, and CSAA Insurance Group are among the carriers using these scores in production. More states are expected to follow as Moody's integrates Cape's data into its catastrophe risk platform.
Cape is not alone in this market. Verisk, EXL, and Nearmap all offer competing aerial analytics products that collectively mean your property is being scored by multiple independent algorithms simultaneously, any one of which could trigger an underwriting action. Verisk's Roof Condition Score is now a core underwriting signal for carriers nationwide. EXL combines roof condition with fire risk scores, flood risk scores, and liability detection for swimming pools, trampolines, and fences. An algorithm can flag your pool and your oak tree in the same pass.
What a Bad Score Costs You
Start with premiums. Verisk's 2026 report documents that the premium gap between homes with roofs under five years old and those aged 11 to 15 years expanded from $49 in 2022 to $155 in 2025, representing a 216 percent increase in three years, driven almost entirely by the adoption of aerial imagery scoring that allows insurers to price roof age with a precision that simply did not exist when the field inspector visiting once every few years was the industry standard for property assessment.
Layer on deductibles, which rose 22 percent in 2025 after a 15 percent increase in 2024, meaning the deductible you signed up for two years ago in a hail-prone state where the average roof lifespan is 15 years may no longer resemble the deductible on your next renewal in any meaningful sense.
Then consider what happened in March 2026. The Federal Housing Finance Agency announced that Fannie Mae and Freddie Mac will now allow homeowner insurance policies that provide only actual cash value coverage for roofs, rather than replacement cost. ACV means the insurer pays the depreciated value. If your 15-year-old roof costs $10,000 to replace and the insurer calculates its depreciated worth at $2,000, you pay the remaining $8,000 out of pocket. ACV premiums run 10 to 20 percent cheaper, and many consumers will not realize they have been switched to ACV coverage until they file a claim, because, as Amy Bach of United Policyholders told MarketWatch, some insurance agents "brush off the details."
Put those three layers together and the math is punishing.
| Cost Layer | Impact | Source |
|---|---|---|
| Premium penalty (poor score vs. good) | $155+/year, expanding | Verisk 2026 Roof Report |
| Deductible inflation | +22% in 2025 alone | Verisk 2026 Roof Report |
| ACV coverage gap on $10K roof | Up to $8,000 out of pocket | FHFA/Fannie Mae, March 2026 |
| Non-renewal risk (worst states) | Up to 136% higher since 2018 | U.S. Senate Budget Committee, Dec 2024 |
Over a 10-year ownership period on a home with a poor AI roof score in a state with rising non-renewal rates, the compounding financial exposure easily reaches five figures, a penalty assessed by an algorithm you cannot see, on a timeline you cannot control, under rules that vary wildly depending on which side of a state line your house happens to sit on.
The Regulatory Map Has Gaping Holes
Connecticut got there first. Its Insurance Department issued guidance advising carriers that aerial images showing discoloration, streaking, or other cosmetic issues alone should not support adverse underwriting action. Insurers must rely on clear indicators of actual roof condition consistent with their filed guidelines, not algorithmic impressions of what moss looks like from space.
New York has a bill pending that would go further: require date-stamped images, prohibit reliance on imagery older than 180 days, mandate a 60-day remediation window before cancellation or non-renewal, and create an appeals process that includes the option of an in-person inspection when imagery is inconclusive. Massachusetts requires 45-day advance written notice with specific reasons for non-renewal. Ten states require insurers to offer discounts for roof upgrades to wind-resistant standards.
Nationally, the NAIC's Model Bulletin on the Use of Artificial Intelligence Systems by Insurers provides the governance framework. Its core principles: do not run AI models on bad data, validate image quality and recency before using it in underwriting, and do not outsource accountability to vendors. A computer vision model that systematically downgrades older roofs in certain neighborhoods or misreads shadows as structural damage can create systematic disparities that would violate fair underwriting standards if a human examiner did the same thing.
But the NAIC Model Bulletin is guidance, not law. Adoption is voluntary. As of July 2026, most states have neither adopted it nor enacted their own aerial imagery regulations. If you live in Texas, where non-renewal rates climbed 1.96 percent between 2018 and 2023 alongside average premiums of $6,005 per year, no state law requires your insurer to show you the satellite image that triggered your non-renewal, give you time to fix what the AI flagged, or explain whether the algorithm confused a shadow with a shingle crack.
Builders Should Pay Attention
If you are building homes in 2026, roof material selection just acquired a second audience: the algorithm. Impact-resistant shingles rated Class 4 under UL 2218, hip roof geometries that reduce wind exposure per square foot, proper drip edge installation, and light-colored roofing materials that minimize thermal contrast in satellite imagery all influence AI scoring. A builder who installs a dark-colored three-tab shingle on a complex gable roof with no drip edge and a couple of overhanging trees has inadvertently built a home that starts its insurance life with a mediocre AI score, regardless of how well the installation was done.
For production builders, the economics are becoming measurable. A home that photographs well from orbit faces lower buyer insurance costs, which improves affordability ratios and reduces deal friction at closing. A home that scores poorly from the first satellite pass creates a liability that follows the buyer until they replace the roof.
Brickeye, a Toronto-based sensor company that raised $10 million in a January 2026 Series B, is pushing this logic into the construction phase itself. Their monitoring systems document water management, curing conditions, and environmental controls during construction. CEO Tim Angus told ENR that "insurers are no longer satisfied with general statements about risk management. They want to understand what controls are actually in place on the jobsite and how losses are being prevented in real time." Builders' risk underwriting, which covers the construction period, is following the same trajectory as homeowner underwriting: from trust to verification to algorithmic scoring.
In Fairness to the Algorithm
AI roof scoring is measurably more accurate than what it replaced, and acknowledging that matters. For decades, roof condition in underwriting was a function of self-reported age (often wrong), outdated property records (often incomplete), and field inspections that happened once and then aged into irrelevance while the roof itself deteriorated under weather patterns the original inspector never anticipated. Cape's models process current imagery across the entire U.S. housing stock continuously. Verisk's data confirms what the algorithms detect is real: roofs scored moderate-to-poor carry 60 percent higher loss costs. Poor roofs do cost insurers more, and they cost homeowners more too, because a 15-year-old roof with missing shingles genuinely fails more often in a windstorm than one installed last spring.
And the alternative is worse: without aerial scoring, insurers either overprice everyone to compensate for the unknowns or underprice the risky homes and subsidize the losses with premiums from well-maintained properties, and neither outcome serves the homeowner who actually takes care of their roof. AI scoring creates the possibility of precision pricing: you maintain your roof well, you pay less. That is a better system, in theory.
But theory requires transparency to become practice, and a score you cannot see, generated by a model you cannot interrogate, applied under rules that vary by state and by carrier, is not precision pricing so much as an opaque penalty imposed without an appeals process. When that penalty compounds with ACV coverage shifts and rising deductibles, the homeowner with a 12-year-old roof and a moss patch faces a financial exposure that the polite phrase "premium adjustment" does not begin to describe.
What You Should Do Right Now
If you own a home: request your insurer's underwriting file. Ask whether aerial imagery or AI-derived property attributes were used in your most recent renewal decision. If you are in Connecticut, Massachusetts, or a state that has adopted NAIC guidance, your insurer should be able to tell you. If you are in a state with no such requirement, the question itself signals awareness that may influence how your file is handled. Check whether your policy provides replacement cost value or actual cash value for your roof. If you renewed after March 2026 and did not explicitly confirm RCV coverage, read the declarations page line by line.
If you are buying a home: treat the roof's AI insurability the same way you treat the inspection report. Ask the listing agent or seller whether any insurer has non-renewed or declined the property in the past three years, and why. A home with a history of AI-flagged roof issues may cost you thousands more per year in insurance than the listing sheet suggests.
If you are building: spec roofing materials that score well from altitude. Class 4 impact-resistant shingles, hip roof designs, light-colored materials, clean drip edges, and a commitment to keeping overhanging vegetation trimmed are no longer just good building practice. They are inputs to an algorithm that will assess your buyer's insurance costs on the day the satellite next flies over.
Limitations
No public data isolates how many homeowner non-renewals are specifically triggered by AI aerial imagery versus other underwriting factors such as claims history, catastrophe zone exposure, or roof age alone. Cape Analytics' and Verisk's scoring methodologies are proprietary, and no independent third-party audit of their accuracy or bias has been published. Regional premium differentials cited in this article reflect aggregate data that does not control for all variables affecting homeowner insurance pricing. New York's proposed aerial imagery legislation has not been enacted as of this writing. ACV impact calculations use simplified depreciation. Real insurer calculations are opaque and vary by carrier, policy, and state regulatory environment.