Your Insurer's AI Already Scored Your Home for Wildfire Risk. You Weren't Told.
ZestyAI analyzed 11.8 million California properties and found 1.2 million homes the algorithm rates as high risk that FEMA rates as low or none. The construction features that move your score cost $2,800 to address. Whether that investment pays for itself in premium reductions depends on math most homeowners have never been shown.
Somewhere in a data center, a machine learning model is staring at a satellite photo of your roof. It knows the species of the tree touching your eave. It can see whether that fence connecting to your neighbor's yard is wood or metal, and it measured the distance from your deck to the nearest patch of dead brush across 11.8 million California residential properties over four consecutive years of annual high-resolution imagery, cross-referencing vegetation density, terrain slope, building characteristics, climate factors, and proximity to more than 2,000 historical wildfire perimeters to produce a risk assessment that no human reviewer, no matter how experienced, could replicate at anything approaching that scale or consistency.
It gave your home a number, and your insurer used that number to decide whether to keep you. Nobody told you.
ZestyAI published those numbers in January 2026, and the headline figure is staggering: 1.2 million California homes that FEMA classifies as low-risk or gives no wildfire rating at all are flagged as high-risk by the Z-FIRE model. Those homes represent $940 billion in property value, with a median price of $800,000 per unit. Among them, 300,859 were built before 1980, predating every modern fire-resistant construction standard on the books.
FEMA, it turns out, simply hasn't looked at most of them. A full 62.5% of California residential properties carry no federal wildfire assessment whatsoever, and insurers filled that classification void themselves using proprietary AI models that homeowners cannot see, cannot challenge, and cannot meaningfully respond to even if they knew the scores existed.
What the AI Sees That FEMA Doesn't
The Z-FIRE model produces two scores: an L1 hazard rating measuring the likelihood that wildfire reaches your property, and an L2 destruction score estimating the probability that fire actually destroys the structure if it arrives. This distinction matters enormously for construction, because L2 is where your building materials, defensible space, and hardening investments show up in the math.
ZestyAI tracked 9.27 million California properties from 2022 to 2025. Among homes scoring 7 or above on hazard, 41.3% reduced their destruction risk through visible mitigation efforts detected by satellite imagery. The AI noticed when homeowners cleared vegetation, replaced roofing, or created ember-resistant zones around their foundations, updating its scores accordingly in the next annual pass with a granularity and consistency that the federal system, which showed what ZestyAI characterized as "little change" over the same four-year window, could not approach.
But the bad news is worse than the good. Among properties scoring 9 or above on hazard, 52% saw their regional wildfire probability actually increase, with an average 1.3-point score jump translating to a 51% rise in annual fire likelihood. In the Palisades fire zone, ZestyAI had flagged 1,430 properties as elevated risk that FEMA classified as low. Value at stake: $1.14 billion. In the Eaton fire zone, 1,615 properties sat in census tracts FEMA rated "Very Low" or gave no rating at all. Value at stake: $1.29 billion. Both fire zones proved the AI right.
More than a third of the California insurance market now relies on Z-FIRE scoring, and the California FAIR Plan, the insurer of last resort, signed a four-year partnership expansion with ZestyAI. This is the scoring infrastructure that determines whether you can get coverage at all.
The $2,800 Question
Here is where construction meets insurance in a way that most builders haven't absorbed yet, and the answer is surprisingly cheap.
Headwaters Economics, working with the Insurance Institute for Business & Home Safety, calculated the marginal cost of three wildfire resistance levels for new California residential construction. Baseline means Chapter 7A compliant, the current minimum code. Enhanced adds $2,800 over baseline. Optimum adds $18,200 to $27,100 depending on whether you're in northern or southern California.
That $2,800 gets you a vertical under-deck enclosure and a noncombustible zone within five feet of the foundation, earning the IBHS "Wildfire Prepared Home" designation. On a typical California new-build where total construction costs range from $350,000 to $600,000 depending on size and market, the Enhanced package represents somewhere between 2% and 8% of the budget, which is the kind of line item that vanishes into the noise of a construction pro forma while potentially determining whether the completed home is insurable at all in five years.
Now connect those two datasets. The AI scoring infrastructure rewards specific, satellite-visible construction features. California Regulation 2644.9 mandates that insurers provide wildfire risk scores to policyholders and offer discounts for hardening. The state's "Safer from Wildfires" framework lists ten specific steps, each of which qualifies for a discount: Class-A fire-rated roof, five-foot ember-resistant zone using stone or decomposed granite, ember-resistant vents with 1/16- to 1/8-inch metal mesh, noncombustible material for the first six inches of exterior walls, enclosed eaves with soffits, multi-paned windows, cleared vegetation under decks, combustible outbuildings moved beyond 30 feet, defensible space compliance, and participation in a community program like Firewise USA.
The AI can detect all of them from aerial imagery. The regulatory framework rewards all of them with mandatory premium discounts. Run the math. If your annual wildfire surcharge exceeds $1,400, the Enhanced upgrade pays for itself within two years. In fire-prone zip codes where surcharges routinely hit $3,000 to $8,000 annually, payback drops below twelve months.
Physics-Based Models Are Getting More Specific
Stand Insurance, founded by former Metromile CEO Dan Preston with $65 million in combined funding from Inspired Capital, Eclipse, Lowercarbon, and others, goes further than statistical scoring. Stand builds physics-based AI simulations of how fire would actually burn on your specific property. The model identifies individual tree species, calculates burn temperatures per species, and maps flame exposure vectors to specific building surfaces.
Instead of generic guidance like "create defensible space," Stand tells you which window to replace. Specifically: that eucalyptus 12 feet from the southeast corner of your house burns at a temperature that exceeds the thermal rating of single-pane glass but falls below the threshold for dual-pane tempered, so you replace that one window, leave the others alone, and the model recalculates your fire exposure profile with the specificity of an engineering analysis rather than the broad strokes of a zone-based recommendation. Premium reduction: potentially up to 60% in fire-prone regions for homes paired with Frontline Wildfire Defense's active suppression systems, which qualify for an additional 25% discount.
On the consumer assessment side, WyldSafe, founded by Woolsey Fire survivor Nate Siggard, launched a patent-pending computer vision app that lets homeowners walk their property with a phone camera. The AI analyzes every frame for combustible vegetation, vulnerable roof materials, unscreened vents, wood fencing, and ignition points. It delivers risk ratings, repair priorities, estimated costs, and a suggested timeline. It is free to homeowners, with costs covered by a network of mitigation vendors. WyldSafe timed the launch to coincide with California's AB 888 Safe Homes grant program, which allocates public funds for exactly this kind of property-level wildfire hardening.
What This Means If You're Breaking Ground
Spec the Enhanced package from day one. Do it. The $2,800 marginal cost is the cheapest insurance investment on the entire project. Noncombustible five-foot perimeter zone using pea gravel instead of bark mulch, enclosed eaves, ember-resistant vents, composite decking instead of wood. Every single item is visible from satellite imagery, which means the AI scoring model will detect the changes on its next annual pass, which means the regulatory discount framework mandated by Regulation 2644.9 will apply to them automatically without you needing to call anyone or file anything beyond what the insurer's own aerial assessment captures. Do not wait until the insurer asks.
Request your Z-FIRE score before you close on land. California Regulation 2644.9 now requires insurers to provide wildfire risk scores to policyholders. First Street Foundation's Fire Factor scores are already embedded in Realtor.com and Zillow listings, using 100 million simulated fire events to produce a 1-to-10 risk rating for every US property based on construction type, roof type, terrain, weather, and proximity to natural fuels. If you're evaluating a lot, that score tells you more about future insurance costs than anything the listing agent will volunteer.
Understand that "insurable" is now a construction specification. This is the shift. Stand Insurance aims to write $2 billion in home coverage within its first year by insuring properties too valuable for the FAIR Plan but abandoned by conventional carriers. The path to coverage runs through specific, AI-verifiable construction features. Your architect and your general contractor both need to understand this. The window schedule, the vent specification, the siding material, the landscaping plan within the first 30 feet of the structure: all of these are now insurance inputs, not just aesthetic or code-compliance choices, and the difference between a home that costs $4,000 annually to insure and one that no carrier will touch at any price may come down to whether the deck boards are composite or wood and whether the eave vents use 1/8-inch mesh or 1/4-inch mesh.
Budget $800 to $2,000 for a professional wildfire risk assessment if you already own a home. WyldSafe is free but unproven. A Firewise USA site assessment from a qualified wildfire mitigation specialist will map every vulnerability on the property, prioritize repairs by cost-effectiveness, and generate documentation that satisfies both the state's Safer from Wildfires framework and your insurer's discount qualification process. AB 888 grants may cover a portion of the assessment and subsequent hardening costs. Do not rely on a vendor-funded free app as your sole assessment when $800,000 in property value and $5,000 in annual premiums are at stake.
Limitations of This Analysis
The $2,800 Enhanced cost figure comes from Headwaters Economics using RSMeans national construction cost estimates, which may understate current California labor and material costs by 15% to 30% depending on region and trade availability. The payback calculation assumes wildfire surcharges of $1,400+ annually, which applies in high-risk zones but not everywhere in the state; in moderate-risk areas, payback stretches to three to five years. ZestyAI's analysis uses its own proprietary model to critique federal risk maps, creating an inherent conflict of interest that merits disclosure even if the underlying data appears sound. Stand Insurance's "up to 60%" premium reduction claim has not been independently verified by any third-party actuarial review. WyldSafe is pre-revenue with no published accuracy metrics for its computer vision model. No longitudinal study has established whether AI-recommended construction mitigations actually reduce fire loss rates in practice, because the scoring systems haven't existed long enough to generate statistically significant post-fire performance data. Finally, 71% of recent homebuyers told the National Association of Realtors they consider natural disaster risk when choosing a location, but property-level AI scoring can create a form of algorithmic redlining if it systematically devalues homes in lower-income communities that lack resources for hardening investments, a concern no current regulatory framework adequately addresses.