Sometime in the last 18 months, an algorithm looked at your house from orbit and assigned it a score. It evaluated your roof condition, your tree coverage, your driveway, the debris in your yard, and whether that new shed you built last summer actually matches the footprint on file. It did this without knocking, without calling, without sending a letter. That result went straight into your insurer's underwriting system, where it influenced your renewal, your premium, or your cancellation notice.
You were not consulted.
How the Scoring Works
Cape Analytics, which Moody's acquired in January 2025, is the company most insurers won't name when they cancel your policy. It runs computer vision models on satellite and aerial imagery for over 80 enterprise clients, nearly all of them insurance carriers, producing what the industry politely calls "property intelligence." Its system ingests imagery from providers like EagleView and Nearmap, layers in weather data and public records, and outputs risk assessments granular enough to tell your underwriter whether that dark patch on your roof is a shadow or a problem.
What does it score? Roof condition and estimated age. Building footprint and number of stories. Outbuildings, additions, pools, solar panels. Vegetation proximity and defensible space in wildfire zones. Hail damage indicators. Debris accumulation. Staining, discoloration, and what the system interprets as wear.
That last category matters enormously. The distinction between "your roof has structural damage" and "your roof has cosmetic discoloration from algae" is the difference between a legitimate underwriting concern and a false alarm. The AI does not always know which is which.
When the Algorithm Gets It Wrong
A March 2026 analysis in Carrier Management by Troutman Pepper Locke attorney Paige Waters cataloged the regulatory response. The concerns are specific: "structures may be obscured by trees or shadows, misaligned, or even mis-identified." A computer vision model that reads a shadow as a hole, or interprets dark algae streaking as missing shingles, produces a score that looks authoritative to the underwriter who never visited the property.
State regulators have started drawing lines. Several insurance department bulletins now distinguish between conditions that "materially increase the risk of loss" and "cosmetic roof or property conditions, such as staining, streaking, discoloration or minor aesthetic flaws." In plain language: your roof looking old is not the same as your roof being dangerous. The algorithms conflate the two more often than the industry admits.
An Insurance Research Council survey found that 31% of homeowners cite accuracy as their biggest concern about aerial imagery in insurance, making it the number one worry. Sixty percent of homeowners are aware their insurer uses this technology. The remaining 40% are being scored without knowing it.
What This Means If You're Building New
New construction already carries a massive insurance advantage. A two-year-old home with $300,000 in dwelling coverage averages $1,468 per year in premiums, compared to $2,276 for a 20-year-old home. That is 35% less, or $808 annually, according to Hotaling Insurance Services. Modern electrical systems alone reduce fire risk by 60% versus pre-1980 wiring.
But the aerial scoring adds a new variable. Your construction choices now feed two separate evaluation systems: the traditional underwriting model (which cares about building codes, materials, and claims history) and the aerial AI model (which cares about what your property looks like from 500 feet up). They do not always agree.
A concrete tile roof in a color that photographs poorly from above could score worse than a lighter asphalt shingle roof that reflects more light, even if the tile roof is objectively more durable and fire-resistant. A vegetated roof designed for stormwater management might trigger a "debris accumulation" flag. Solar panels, depending on their arrangement and the model's training data, can confuse footprint calculations.
Construction Choices With the Highest Insurance ROI
Start with the roof. Class 4 impact-resistant shingles carry an additional 10% to 28% insurance discount depending on the state and carrier, according to industry rate filings. More importantly, they score well on aerial assessment because they maintain their visual integrity longer. A roof that still looks crisp and uniform from above at year eight does not trigger a condition flag the way a standard architectural shingle with early granule loss does.
Then there is IBHS Fortified designation. In Alabama, the only state with regulated minimum discounts, Fortified Gold homes receive up to 55% off the wind portion of premiums. Fortified Roof designation gets 35% off. Louisiana has issued over 5,400 Fortified certificates as of early 2025, a seventeen-fold increase since 2023, with homeowners reporting median annual savings of $1,250, a 22% reduction.
Fortified math is stark. After Hurricane Sally, researchers at the Center for Risk and Insurance Research estimated that if all conventional homes in the affected area had met Fortified Roof standards, policyholders would have reduced their costs by 66%. At Fortified Gold, the savings reached $34.6 million in deductibles alone, a 65% reduction.
| Construction Choice | Aerial AI Impact | Insurance Savings |
|---|---|---|
| Class 4 impact-resistant shingles | Maintains visual uniformity longer, fewer condition flags | 10-28% discount |
| IBHS Fortified Roof | Verified structural resilience, positive risk score | Up to 35% off wind premium |
| IBHS Fortified Gold | Full-envelope hardening, highest resilience tier | Up to 55% off wind premium |
| Light-colored roofing | Photographs clearly from above, fewer shadow-induced false flags | Indirect: avoids non-renewal risk |
| Defensible space (wildfire zones) | AI measures vegetation proximity to structure | Varies by carrier, often required for coverage |
| New construction (any material) | Clean imagery baseline, no accumulated wear signals | 33-35% lower premiums vs 20-year homes |
Discount percentages sourced from state insurance department rate filings (Class 4 shingle discounts), IBHS Fortified program data via Epicenter Insights (Fortified tiers), and Hotaling Insurance Services analysis (new construction advantage). Aerial AI impact descriptions are inferred from Cape Analytics' published scoring categories and industry documentation, not from the company's proprietary model weights.
Your Rights When the Score Is Wrong
Regulators are catching up, unevenly. The NAIC AI Model Bulletin establishes that insurers cannot outsource accountability. If their vendor's aerial AI produces a bad score, the insurer is responsible for the downstream decision. State bulletins are layering specific requirements on top: insurers must validate image quality and recency before using it in underwriting, especially when AI scoring tools are involved.
California has proposed measures requiring insurers to notify homeowners before using aerial imagery and to provide copies of images on request. Pennsylvania's insurance department has reminded carriers that aerial imagery alone may not meet the legal threshold for a "significant change in risk" required to justify non-renewal. Several states now expect insurers to give homeowners an opportunity to make repairs before finalizing a cancellation.
If your insurer sends a non-renewal notice citing property condition, request the specific imagery and scoring data. Present a contractor's report or recent inspection if the AI's assessment is wrong. Regulators increasingly expect insurers to conduct a follow-up physical inspection when a homeowner disputes an aerial finding. The algorithm's opinion is not supposed to be final. In practice, whether it is depends on your state and your willingness to push back.
In Defense of the Eye in the Sky
Before aerial AI, insurers priced residential policies using self-reported data from homeowners (frequently wrong), publicly available property records (frequently outdated), and expensive in-person inspections that happened rarely, if at all. Most policies were priced using the age and location of the home plus claims history and almost nothing about its current physical condition. Aerial scoring, for all its misidentification problems, replaced that near-total blindness with actual observation. It catches tree limbs resting on rooflines, debris-choked gutters, and collapsing outbuildings that would have gone unnoticed until they caused a claim. According to the IRC survey, 55% of homeowners believe aerial imagery leads to fairer pricing, which is not nothing. Regulators' objections center on execution, not the concept. Nobody is arguing insurers should go back to guessing.
Limitations
Cape Analytics does not publicly disclose false positive or false negative rates for its roof condition scoring. The IRC homeowner survey was funded by the Insurance Research Council, an industry group, which introduces potential response bias. IBHS Fortified discount data is concentrated in hurricane-prone coastal states; inland savings are less dramatic and vary by carrier. State regulatory responses range from detailed bulletins with enforcement mechanisms to no guidance at all. We lack systematic data on what percentage of non-renewals are driven specifically by aerial AI versus other underwriting factors.
Catherine Chen covers policy and regulation for AI Home Building. She finds it remarkable that an industry built on actuarial precision is comfortable making coverage decisions based on what a satellite saw through a cloud on a Tuesday.