Your Insurer's AI Already Knows Your House Will Burn. The Fix Costs $9,000.
Machine learning models now predict which homes survive California wildfires with 82% accuracy. Building a fire-resistant home costs 2% more. The gap between what AI knows and what we do about it is where $30 billion in losses lives.
In Eagle Rock, a neighborhood perched in the hills of northeast Los Angeles, a two-bedroom ADU went up in ten days this spring: insulated concrete form walls, cast-iron plumbing vents replacing the standard plastic ones that melt and open a fire pathway straight into the attic, fire-resistant windows. Total construction cost: $410,000, which is roughly what the homeowner was quoted for a conventional wood-framed unit. It is, according to Builtech Construction, the first home in Los Angeles built to Type I standards under the International Building Code, the highest fire-resistance classification that exists. It sits minutes from the Eaton Fire evacuation zone, where 7,000 structures burned to foundations in January 2025.
That gap between what's possible and what's standard is where $30 billion in losses lives.
A study published in Nature Communications last August, led by UC Berkeley mechanical engineering professor Michael Gollner and postdoctoral researcher Maryam Zamanialaei, trained a machine learning model on CAL FIRE's damage inspection database, which contains on-the-ground surveys of every structure damaged or destroyed in major California wildfires since 2013. Tested against five catastrophic fires, including the 2017 Tubbs and Thomas fires, the 2018 Camp Fire, the 2019 Kincade Fire, and the 2020 Glass Fire, it predicts whether a given structure will survive or be destroyed with 82% accuracy.
Two findings stood out: combining home hardening with defensible space doubled survival rates, cutting predicted wildfire destruction by 50%, and just clearing vegetation within five feet of a structure, the regulation California calls Zone 0, reduced structure losses by 17% on its own.
That second number deserves to sit with you. Seventeen percent fewer homes destroyed, and the intervention is pulling out the juniper bushes touching your siding.
What Insurers Already Figured Out
While homeowners debate whether to spend $300 on ember-resistant vent screens, insurers have been scoring their homes with AI for years.
ZestyAI's Z-FIRE model covers approximately 100% of U.S. properties. It analyzes more than 2,000 historical wildfires using satellite and aerial imagery, topography, and property-level characteristics to generate parcel-specific wildfire risk scores. More than a third of California's insurance market uses it, including the California FAIR Plan, which is the state's insurer of last resort for the 788,000 homeowners who have been dropped by private carriers. California's Department of Insurance has confirmed Z-FIRE as filing-ready for rate segmentation and underwriting without further review.
Verisk completed its own milestone in July 2025: the first wildfire catastrophe model to survive California's rigorous Pre-Application Required Information Determination review, meaning insurers can now, for the first time in the state's history, use a forward-looking, climate-aware wildfire model in their official rate filings. And just this week, Convr announced a partnership with Property Guardian to embed property-level wildfire intelligence, covering structure, parcel, community, and regional analytics, directly into an AI underwriting workbench, reducing the need for in-person loss control inspections.
Delos Insurance, an MGA writing excess and surplus business, partnered with the Spatial Informatics Group, which has done wildfire modeling for Cal Fire and the California Public Utilities Commission for decades, and whose machine learning model has successfully predicted the scope of major wildfires over the past seven years by combining historical geospatial datasets with climate projections and worst-case scenarios. "We innovated using machine learning on geospatial datasets," chief data officer Shanna McIntyre told Insurance Business. "It's not just about historical loss."
The underwriting side of the industry now has property-level AI scoring that can distinguish between the house with dual-pane tempered glass and enclosed eaves and the one with single-pane windows and open soffits sitting on the same street, separated by fifty feet of identical topography, facing the same prevailing winds, priced into entirely different risk tiers. They already know which one will survive and which will burn.
So How Much Does "Surviving" Cost?
Less than your kitchen remodel.
An IBHS and Headwaters Economics construction cost analysis looked at a representative single-family home: one story, 1,750 square feet, in a Southern California setting like Altadena, with an estimated total construction cost around $500,000. They compared costs across three wildfire-resistance standards:
- California Building Code Chapter 7A (now Part 7 of the California Wildland-Urban Interface Code, effective January 1, 2026): adds approximately $13,000, or 2.6% of construction cost.
- IBHS Wildfire Prepared Home: a systems-based standard focused on ember resistance through structural mitigation and defensible space. Adds approximately $9,000, or 1.8%.
- IBHS Wildfire Prepared Home Plus: everything in the base standard plus enclosed eaves and noncombustible gutters. Adds approximately $15,000, or 3%.
Nine thousand dollars. For a half-million-dollar home. Five components carry the weight: roof (covering, vents, edge, gutters), under-eave area (eaves, soffit, vents), exterior walls (siding, windows, doors, trim, vents), attached deck (surface, rails, under-deck footprint), and the first five feet of landscaping around the structure, where RSMeans pricing data, the national database contractors actually use for bidding, anchors every estimate.
Arithmetic Nobody Wants to Do
Here it is anyway.
In 2023, California insurers non-renewed approximately 788,000 homeowner policies, most of them in wildfire-prone areas. At the IBHS Wildfire Prepared Home standard, hardening every one of those homes would cost roughly $7.1 billion. That number assumes the $9,000 new-construction figure, which understates the retrofit cost for existing homes. Call it $15,000 per home for retrofits based on Headwaters Economics' estimate that effective hardening strategies run $2,000 to $10,000, with structural upgrades on the higher end. That puts the total at $11.8 billion for the full 788,000.
January 2025's Los Angeles fires, just two wildfires in one metropolitan area over a few weeks in a single month, caused an estimated $30 billion in losses and destroyed more than 16,000 structures. One event. One city. One month. Two and a half times the cost of hardening every non-renewed-policy home in the entire state.
Anyone with a phone calculator can run these numbers. Spending $12 billion to prevent $30 billion in losses from a single fire season, not counting the 28 lives lost, the 12,900 households displaced, or the cascading economic effects on property values, tax bases, and municipal bonds, is not a policy debate requiring legislative courage or bipartisan coalition-building or any of the other euphemisms politicians deploy when they'd rather not do arithmetic. It solved itself on the back of a napkin.
What the Model Can't Fix
Gollner's Berkeley study identified structure separation distance as the single most influential factor driving loss in densely built areas. When one house ignites, radiant heat and flame contact can set the neighbor's house on fire regardless of how hardened it is. In neighborhoods where houses sit eight feet apart, as in many hillside communities across Southern California, no amount of individual home hardening overcomes the physics of a structure fire six feet from your wall.
This is a planning and zoning problem, not a spending problem. AI models can predict it with precision, but mitigation requires minimum setbacks, building separation standards in fire hazard zones, and decisions that local planning commissions have historically avoided because they reduce buildable square footage and drive up land costs. And that 82% accuracy means roughly one in five predictions is wrong, a gap that matters enormously when you are the one in five. Every fire in the training set occurred before 2022; January 2025's LA fires, driven by extreme Santa Ana winds and record drought, may represent conditions the model hasn't seen.
And then there's the transparency problem, arguably the most intractable piece. ZestyAI, Verisk, and Delos all run proprietary models, which means a homeowner who has been non-renewed cannot independently verify the parcel-level score that determined their fate. California's Safer from Wildfires program promises insurance discounts for every hardening action, but discount amounts are not standardized across carriers. You can spend $15,000 on WFPH Plus certification and still find your premium increase exceeds the theoretical reduction.
What This Means If You're Building
If you are building a new home anywhere in a wildfire-prone zone, and California's new CWUIC Part 7 applies to every new structure in state-designated WUI fire areas as of this year, the premium for fire resistance is 2-3% of your total construction cost. Specify it upfront. The materials aren't exotic. Class A fire-rated roofing, 1/8-inch mesh ember-resistant vents, multi-pane tempered glass windows, fiber cement or stucco siding, enclosed eaves and soffits, noncombustible deck boards within the first five feet of the structure: your contractor already sources every one of these, and the only additional coordination is specifying them on the plans before the first permit drawing goes out.
If you are retrofitting an existing home, start with the $2,000 tier: clear Zone 0, which is the five-foot perimeter of vegetation and combustible material that Gollner's team found responsible for nearly a third of structure losses, then replace standard attic and soffit vents with ember-resistant versions, and remove combustible materials stored against exterior walls. Not one of these actions requires a building permit.
If you are struggling to obtain insurance, ask your agent whether your carrier uses ZestyAI, Verisk, or an equivalent parcel-level model, and whether specific hardening measures would change your risk classification. California's Department of Insurance maintains a Safer from Wildfires checklist, which is not a guaranteed path back to the voluntary market but remains the only structured pathway that exists.
AI already knows which houses will burn, and it has known for years. Whether anyone besides the insurance underwriters will act on what it's saying is the only question left, and at $9,000 per home on new construction, the answer isn't hiding behind a cost problem.
What this article did not prove: Cost figures for the IBHS WFPH standard ($9,000 above traditional construction) apply specifically to new builds in Southern California using RSMeans pricing and a 1,750-square-foot representative home; actual retrofit costs for existing homes range from $2,000 to over $100,000 depending on condition, materials, and scope. Gollner and Zamanialaei's ML model was trained on five pre-2022 California wildfires, and its accuracy on fires driven by different fuel conditions, climate patterns, or urban densities has not been independently validated against post-2022 events. Insurance premium reductions from home hardening are not standardized and vary by carrier, meaning a homeowner who invests in WFPH certification has no guarantee of proportional savings. Statewide hardening cost estimates ($7.1-$11.8 billion) assume uniform per-home costs that would vary significantly by structure age, size, condition, and region. ZestyAI, Verisk, and Delos models are proprietary; independent verification of their parcel-level accuracy is not publicly available, and the homeowner's ability to contest their score is limited.