AI in Homeowners Insurance Pricing and the State Rate Review

How homeowners insurers use AI, aerial imagery, and catastrophe models to price property risk, and what state regulators now require for transparency.

Homeowners insurance pricing has become one of the most visible AI battlegrounds in insurance. For decades, rates were set by territory, construction type, age of home, and claims history. Underwriters might inspect a property every few years. Today, carriers can price a home by combining satellite imagery, roof-age models, wildfire exposure scores, and climate-adjusted catastrophe models, sometimes without ever sending a person to the property. The process is faster and more granular. It is also more opaque to the homeowner and more likely to produce non-renewals or large rate increases.

The market pressure is real. The average premium for a new home insurance policy reached $1,952 in December 2025, up 8.5% year over year, though that was a slowdown from the 18% increase seen between 2023 and 2024 ¹. Insurance now accounts for an estimated 9% of the typical homeowner’s monthly mortgage payment, the highest share on record ¹. In high-risk areas, the cost and availability of insurance are beginning to affect home values. Homes in the top 25% of hurricane and wildfire exposure have lost roughly $20,500 in value since 2018, and those in the top 10% have lost roughly $43,900 ¹.

AI is not the only cause of these trends, but it is the tool that makes them precise.

What AI sees when it prices a home

The inputs used in AI-driven homeowners pricing fall into four categories.

Aerial and satellite imagery is the most visible. Vendors capture high-resolution images of nearly every property in the country and use computer vision to identify roof condition, roof age, tree overhang, yard debris, pool, trampoline, and other property features. In 2024, U.S. roof claims costs reached nearly $31 billion, up about 30% since 2022, which has made roof condition one of the most influential underwriting factors ¹.

Catastrophe and climate risk models estimate the probability of wildfire, hurricane, hail, and flood at the property level. These models are grounded in historical event data and climate science, but they are also updated frequently as losses change scientific understanding. A carrier that relied on a five-year-old wildfire model may now be exposed to risks it did not price for. The same models, used at treaty level, are discussed in AI in reinsurance treaty pricing and catastrophe modeling.

Property records and permit data include square footage, construction type, year built, and renovation history. These have always been used, but AI can now combine them with imagery and claims data to produce more granular risk scores.

Third-party risk scores combine multiple data sources into a single score that influences pricing, acceptance, or renewal. The score may be proprietary, and the carrier may not fully understand how it is built. That is where the governance problem begins.

The pricing and coverage consequences

The practical result of these tools is that carriers can identify individual properties that no longer fit their risk appetite and move them out of the book. This shows up in three ways: higher premiums, higher deductibles, and non-renewals.

Premium growth has been highest in catastrophe-prone states. In California, thousands of homeowners saw renewals rise 200% to 300% in a single year due to AI-derived risk updates, with similar properties nearby sometimes escaping the same increases ². The selectivity is the point of granular risk scoring, but it is also what makes the process feel arbitrary to the consumer.

Deductibles are rising as a separate affordability tool. The average home insurance deductible rose 22% in 2025, after a 15% increase in 2024 ¹. A higher deductible lowers the premium but shifts more cost to the homeowner at the time of a claim. It is a way to keep the monthly payment manageable while reducing the carrier’s exposure.

Non-renewals are becoming more common in high-risk areas. A carrier may use aerial imagery to flag a roof, tree overhang, or yard debris as a reason not to renew. In some cases, the images have been outdated or inaccurate. A California Department of Insurance investigation found cases where imprecise drone or satellite photos led to wrongful non-renewals ³.

The regulatory pushback

State regulators are responding with two tools: transparency requirements and rate review scrutiny.

California has been the most aggressive. In March 2025, Insurance Commissioner Ricardo Lara announced support for Assembly Bill 75, which would require insurers to notify homeowners at least 30 days before obtaining aerial images of their property and would give homeowners the right to request and obtain copies of those images ³. The bill is aimed at the core problem of AI-driven homeowners pricing: homeowners do not know what data is being used, whether it is accurate, or how to challenge it.

Other states have taken different approaches. Colorado, Georgia, and Kentucky require aerial imagery used for underwriting to be no more than 12 months old. Rhode Island allows 15 months ⁴. Colorado’s broader automated decision-making framework also imposes disclosure and appeal requirements on insurance AI; see the analysis of Colorado SB 26-189. These rules are narrower than California’s notice-and-access proposal in the imagery-specific context, but they address the same concern: a model is only as good as the data it uses, and stale or inaccurate imagery can lead to wrongful pricing or non-renewals.

FIG. 1 — STATE LIMITS ON AERIAL IMAGERY AGE FOR UNDERWRITINGSOURCE: NEARMAP, INSURANCE REGULATIONS BY STATE, 2026

Rate review scrutiny is the broader regulatory lever. The NAIC notes that state insurance regulators oversee insurers’ use of AI and may require companies to explain how these tools are used in underwriting, pricing, marketing, or claims decisions ⁵. In homeowners, that explanation increasingly becomes part of the rate filing. A state department of insurance can ask whether the aerial imagery, wildfire score, or catastrophe model output is actuarially justified, whether it is unfairly discriminatory, and whether it is based on accurate data.

The governance challenge

For carriers, the challenge is not whether to use AI in homeowners pricing. It is whether the AI can survive a rate filing review, a consumer complaint, or a bad-faith lawsuit.

The first control is data accuracy. A carrier must know the age of the aerial imagery, the source of the roof-age estimate, and the vintage of the catastrophe model. Imagery that is more than a year old should be refreshed before it drives a non-renewal. Roof-age estimates derived from imagery should be reconciled against permit data or inspection reports where possible. These are the same data-ownership and freshness questions the AI inventory by line of business playbook asks carriers to record.

The second control is model explainability. A wildfire score or catastrophe model output must be translatable into an underwriting or pricing rationale that a regulator can understand. If the model says a property is high risk, the carrier should be able to explain which inputs drove that conclusion and why they are related to the peril being priced. The governance framework for this is covered in AI governance in insurance.

The third control is human review. AI can flag properties for non-renewal or rate increase, but a human should review the decision before it is finalized, especially when the action is adverse to the consumer. The human reviewer should have access to the underlying data, the ability to override the model, and a system that records the reason for the override.

The fourth control is consumer transparency. Homeowners should be told, in plain language, what data is being used to price or renew their policy and how to dispute it. California’s AB 75 is an early version of this requirement, but the direction is national. Regulators are moving from the question of whether AI can be used to the question of whether consumers can see and challenge what it does.

Where this fits in the broader risk map

Homeowners pricing is the consumer-facing edge of AI in P&C. The same aerial imagery, catastrophe models, and third-party scores are used in commercial property and reinsurance, but homeowners is where the backlash is loudest because the policyholder is a retail consumer who can vote and complain to the regulator. For the broader picture of where AI is used across lines, see AI use cases in insurance by business line.