AI is changing how P&C insurers model climate risk

• Climate peril severity at highest level in seven years
• Recent lower frequency of perils won't last
• AI models getting better at evaluating visual property risk information

The rising severity and frequency of climate peril are pushing insurers to apply AI and other new technologies to modeling and mitigating the risks, data analytics providers say.
Severity rose significantly for all perils, to the highest it's been over the last seven years, increasing almost 26% from 2024 and over 93% compared to 2019, according to LexisNexis's latest home trends report. 

While loss costs across all perils dropped by over 4% in 2025 and frequency decreased nearly 24% from the prior year, LexisNexis analysts note that the drop in claims frequency should not be expected to continue, as the number and severity of billion-dollar weather events and storms rise.

The key to keeping peril costs in check despite an overall upward trend is how carriers apply technology to the problem, said George Hosfield, vice president and general manager of home insurance at LexisNexis Risk Solutions. Because of the seasonal and year-to-year variability of perils, it is essential for carriers to understand both by-peril and macro-level trends through more granular property intelligence and technology, Hosfield said. 

AI models using visual language are improving digital self-inspection tools for evaluating home preparedness and risk-hardening measures, according to Hosfield and Kelly Rush, director of home insurance at LexisNexis Risk Solutions. These models interpret image-based information and apply it to risk evaluation.

"Technology is increasingly being used to identify property-level issues before a loss occurs," Hosfield and Rush wrote in an email response to questions. "Digital self-inspection gives insurers a more cost-effective way to gather detailed information about a property's condition and surrounding risk factors."

AI is becoming more suited to evaluating visual information about property risk, according to Chris Lucas, principal machine learning engineer at Fathom, a climate risk data provider specializing in flood data.

"Machine learning AI is very good at spotting patterns that we wouldn't be able to spot," he said. "If we give it lots of examples of terrain, it would be able to tell the difference in ways in which we simply couldn't. It enables us to zoom in on areas and produce much higher resolution simulations."