Linda Bennett had been a State Farm customer for twenty years when the letter arrived. It demanded $20,000 in roof repairs. No inspector had visited her Santa Ana home. No one had knocked on her door, climbed a ladder, or pressed a thumb into a suspect shingle. State Farm confirmed to GadgetReview that it uses "a mix of aerial images" from drones, satellites, and manned aircraft, all processed through AI systems that generate a roof condition score the homeowner never sees, never consents to, and in Bennett's case, had no visible damage to justify.
She is not unusual. She is the process working exactly as designed.
How the Pipeline Works
I have managed construction projects for two decades, which means I have spent two decades watching invisible systems quietly determine whether someone's project goes forward or dies in a filing cabinet. Insurance underwriting has always been one of those systems. What changed is that the human adjustor who used to drive out, park in your driveway, squint at your gutters, and make a judgment call has been replaced by a pipeline that processes your roof from 400 feet in the air without telling you it happened.
The architecture is straightforward, and that is part of what makes it so difficult to fight. CAPE Analytics sells a Roof Condition Rating product that uses computer vision to assess degradation from aerial photos. Their Roof Age tool claims 95% accuracy by combining historical aerial imagery with permit data. Nearmap (which acquired Betterview) runs AI roof age estimation with 3D mesh technology. EagleView's Assess product deploys drones for roof inspection, already adopted by Hastings Mutual. Verisk publishes a Roof Underwriting Report combining high-resolution imagery with AI analysis. Behind all of them sits the Geospatial Insurance Consortium, a joint venture funded by major carriers, partnered with Vexcel to capture aerial imagery of 99% of U.S. buildings.
That is the supply chain, and your roof enters it without your knowledge, a score comes out the other end, and your insurer acts on it.
What It Costs When the Score Is Wrong
CAPE Analytics claims 95% accuracy on roof age assessment. Take that at face value. A 5% error rate across roughly 140 million housing units in the United States means 7 million homes with potentially incorrect AI roof assessments.
Seven million.
Each false positive triggers a cost cascade that compounds in ways the insurer never has to account for, because the financial consequences flow entirely downstream to the homeowner. An incorrect assessment typically produces one of two outcomes: either the homeowner complies and pays for unnecessary roof work costing $10,000 to $20,000, or the homeowner disputes the finding, loses their coverage, and enters the non-renewal penalty pipeline. Non-renewal follows you everywhere. "A non-renewal can affect your eligibility with other insurers," Katie Walters of Insurify told reporters, which is a polite way of saying that your options narrow to FAIR plans, the insurers of last resort, where premiums typically run two to three times the market rate.
Run the math on a conservative average impact of $10,000 per false positive, covering some combination of unnecessary repairs, premium increases, coverage gaps, and the time cost of fighting a system that was never built to be fought. Seven million affected homes times $10,000 equals $70 billion in potential aggregate homeowner costs generated by AI scoring errors across the U.S. housing stock. Nobody in the insurance industry has published this number, because nobody in the insurance industry bears it.
Cases That Show the Pattern
In Austin, Texas, Tracy Gartenmann received a demand from Travelers Insurance to trim her trees or lose coverage, the basis being grainy aerial photos she had never seen and surveillance she had no idea was happening. In the Houston area, Alaina Callahan's insurer demanded a new roof on a five-year-old roof in demonstrably good condition. She never saw the aerial images used to make that determination. "I had no recourse as a homeowner," she told NPR/KUT investigative reporter Audrey McGlinchy. "None whatsoever."
Twelve Texas homeowners filed state investigations between 2023 and 2025 over aerial photo assessments. Five insurers operating in Texas confirmed using third-party aerial photos, some confirming AI analysis. In one case documented by the NPR investigation, an insurer analyzed the wrong image entirely and dropped the homeowner based on a property that was not theirs.
Wrong image. Wrong house. Coverage gone.
Amy Bach, executive director of United Policyholders, a consumer advocacy organization, described the pattern clearly: AI "often misinterprets images, leading to overreactions like dropping long-term policies despite no actual changes." Mark Friedlander of the Insurance Information Institute offered the industry's rebuttal with equal clarity: "Companies that still haven't adopted this technology? They will eventually."
So this scales.
Non-Renewals Are Accelerating Everywhere
According to the U.S. Senate Budget Committee's December 2024 report, non-renewal rates have exploded across the country: Florida saw a 280% increase between 2018 and 2023, Louisiana 267%, Hawaii 216%, South Carolina 136%, Oklahoma 103%, and California 82%. Weather catastrophes drove 97% of insured losses worldwide in 2024, totaling $140 billion per Munich Re. Home insurance rates rose 10.4% in 2024 alone, and Insurify projects the average U.S. premium will reach $3,057 by 2026, though Florida's average already sits at $15,460 per year, with Louisiana close behind at $13,937.
AI aerial assessment did not create this crisis, climate risk did, but AI aerial assessment is the mechanism through which individual homeowners absorb the cost of that risk, often without the ability to see, understand, or challenge the data that triggered their cancellation. Six states experienced premium increases above 20% in 2025. FAIR plans across the country are swelling with displaced policyholders who lost coverage from traditional carriers and landed in the insurance equivalent of the penalty box, paying more for less and carrying it on their disclosure history for years.
California AB 75 and the Regulatory Response
In March 2025, California Assemblymember Lisa Calderon introduced Assembly Bill 75, currently in the Assembly Committee on Privacy and Consumer Protection, which would require insurers to notify homeowners before capturing aerial photos, supply those photos upon request, and limit how images factor into policy decisions. Aerial images older than 45 days could not be used unless the identified problem was persistent and verified through an in-person inspection. If an insurer cancels or non-renews based on aerial photos, they would have to send the images along with an explanation.
California Insurance Commissioner Ricardo Lara stated that his department "has investigated numerous complaints where flawed aerial imagery led to wrongful cancellations or nonrenewals." California already requires 30 days' notice before cancellation, Pennsylvania mandates 60, and Connecticut's Insurance Department issued guidance in February 2026 on how insurers assess roof risk and what rights homeowners retain.
AB 75 would be the most comprehensive aerial surveillance regulation for insurance in the country, and it has not passed. No federal legislation addresses the practice, and most states have no pending bills.
What This Means If You Are Building or Buying
If you are building a new home, your roofing material and color selection now have a second audience beyond the homeowner and the HOA. CAPE Analytics and its competitors are training their models on millions of roof images, and the features those models weigh in their scoring remain proprietary. You cannot design around a scoring algorithm you cannot see. But you can document your roof installation with high-resolution timestamped photos, retain your roofing warranty and permit records, and establish a baseline that gives you evidence to challenge a future AI assessment that contradicts what you installed.
If you are buying, request disclosure of any aerial-imagery-based assessments the seller's insurer performed during the coverage period. That data may not exist in a form the seller can provide, because the seller likely never received it either, but making the request creates a paper trail and signals to the selling agent that you understand the landscape. Factor potential insurance disruption into your carrying cost analysis. If you are buying in Florida, Louisiana, or any state where non-renewal rates have doubled, get an insurance quote before you close. Not after.
If you currently own a home, file a request with your insurer for any aerial imagery or AI-generated assessments associated with your policy, keeping in mind that some states require disclosure while most do not. Document your roof condition independently with professional inspection reports and photographs. If you receive a non-renewal or repair demand based on aerial assessment, contact your state insurance commissioner and United Policyholders before complying with any repair demand, because the assessment may be wrong, and once you pay for unnecessary work, no insurer reimburses you for their algorithm's mistake.
The Strongest Case for the Insurers
Insurance companies are not wrong about the underlying problem: damaged roofs generate the most expensive claims, and deferred maintenance is endemic in residential housing stock. In-person inspections cost $150 to $300 each, cannot scale to 140 million properties, and carry genuine safety risk for the adjustors climbing ladders. AI aerial assessment is faster, cheaper, and does not put anyone on a roof in January.
The efficiency argument is real, and carriers that catch genuine deterioration before a storm season can price risk more accurately, which in theory benefits the entire insurance pool by reducing the cross-subsidization where well-maintained homes pay for the failures of neglected ones. If the technology worked perfectly, and if homeowners had full visibility into the process and meaningful appeal mechanisms, this would be an unambiguous improvement in how residential insurance operates.
It does not work perfectly. Homeowners have no visibility. Appeal mechanisms barely exist. And the financial consequences of errors flow in exactly one direction.
Limitations of This Analysis
CAPE Analytics' 95% accuracy claim is self-reported. No independent audit of any residential AI roof scoring system has been published, which means the 5% error rate used in the $70 billion aggregate cost calculation may be significantly higher or lower than actual field performance. The $10,000 average impact per false positive is a blended estimate across unnecessary repairs, premium increases, and coverage gaps; actual costs vary enormously by market, with a false positive in Miami carrying radically different financial consequences than one in Des Moines. Non-renewal data from the Senate Budget Committee is state-level; county-level variation within those states is substantial. The twelve Texas complaints documented by NPR represent filed state investigations, not total affected homeowners, and the true complaint rate is unknown because most homeowners never file. We cannot isolate how many non-renewals are specifically driven by AI aerial assessment versus other underwriting factors, because insurers do not disclose that breakdown, and we have no mechanism to compel it.