Guides

How insurers use AI in pricing and claims

Where car insurers use AI, from pricing models to photo damage estimates, how state regulators are responding, and what to do if an algorithm affects you.

4AutoInsurance · Published October 4, 2026

Car insurers use artificial intelligence in almost every part of the business: to set prices, decide whom to insure, estimate repair costs from photos, route claims, answer customer questions and flag possible fraud. Insurers remain legally responsible for those decisions, and state regulators increasingly require them to govern and test their AI systems. For drivers, the practical effects are faster simple claims, more data-driven pricing, and a need to check estimates and decisions rather than assume the computer got it right.

Where AI shows up in car insurance

The NAIC's artificial intelligence topic page lists underwriting, pricing, customer service, claims handling, marketing and fraud detection among the areas where insurers use AI, including claims tools that estimate repair costs or assess damage from photos and historical data.

Area What the AI does What you notice
Pricing Models weigh many rating factors and data sources to predict claim costs Prices that differ more finely between drivers and cars
Underwriting Checks applications against outside data and flags inconsistencies Fewer questions, sometimes unexpected follow-ups
Telematics Turns phone or car sensor data into driving scores Discounts or surcharges based on braking, speed and phone use
Claims intake Chatbots and apps collect details and photos 24/7 claim reporting
Damage estimates Image models estimate repair cost or total loss from photos Quick initial estimates, sometimes revised after inspection
Fraud detection Pattern-spotting across claims, networks of providers and repair shops Some claims referred to a special investigations unit

Behavior-based pricing from telematics is the most visible use for drivers; see what driving apps collect and how insurers use it for what those programs collect.

How regulators are responding

Insurance is regulated by the states, and AI rules are developing state by state.

  • NAIC model bulletin. In December 2023 the NAIC adopted a model bulletin on the use of AI systems by insurers. It expects insurers to keep a written AI governance and risk-management program, oversee third-party models and data vendors, and avoid unfair discrimination and other violations of insurance law. As of April 1, 2026, 25 jurisdictions had adopted it, according to the NAIC's adoption map.
  • States with their own rules. The same map lists four states with insurance-specific AI regulation or guidance instead: California (a 2022 bulletin), Colorado (a 2023 regulation, amended effective October 15, 2025), New York (Circular Letter No. 7, issued July 11, 2024) and Texas (a 2020 bulletin).
  • Colorado's law. Colorado's SB21-169, signed in July 2021, requires insurers to test their use of external consumer data, algorithms and predictive models for unfair discrimination against protected classes and to fix harms they find. Private passenger auto insurers that do not use such data must file an annual attestation saying so.

The NAIC's position is that insurers using AI remain responsible for complying with laws on fairness, accuracy and unfair discrimination, and that regulators may require them to explain how the tools are used in underwriting, pricing, marketing or claims.

Photo estimates and claims: what to watch

Photo-based estimating can turn a simple claim around quickly. It can also miss damage that only shows up when a shop removes a bumper or checks sensors. A photo estimate is a starting point, not a final number.

If you file a claim through an app:

  1. Take clear photos in daylight from every angle the app asks for, plus close-ups of all damage and the odometer.
  2. Get your own shop's estimate if the app's number looks low.
  3. Ask the shop to send a supplement when teardown reveals more damage; insurers routinely pay supplements on covered claims.
  4. Ask who reviewed the estimate. You can request that a human claims adjuster look at the claim; see working with an insurance adjuster.
  5. Ask about sensor calibration. Driver-assistance sensors often need recalibration after repairs, and it should be in the estimate; see driver-assistance sensors and the cost of repairs.
  6. Use your policy's dispute options. For value disputes, an appraisal clause may apply. For handling problems, your state insurance department takes complaints.

Pricing and underwriting: your rights

You cannot see an insurer's model, but you have some tools:

  • Adverse action notices. When an insurer charges you more or turns you down based partly on a consumer report, such as a credit-based insurance score or claims history report, federal law requires it to tell you and identify the reporting agency. You can then get a free copy of the report and dispute errors. See credit-based insurance scores and car insurance.
  • Check your reports. Errors in your claims history (CLUE) report or motor vehicle record feed directly into automated pricing. See Your CLUE report and motor vehicle record explained.
  • Ask what data a program uses. Before joining a telematics or connected-car program, ask whether it can raise your price and what happens to the data.
  • Complain when something looks wrong. State regulators can ask insurers to explain how a decision was made.

Benefits and risks in balance

AI can make insurance faster and more accurate: quicker claim payments, prices that reflect how people actually drive, and better fraud detection that keeps costs down for everyone. The risks are errors at scale, outside data sources that act as proxies for protected characteristics, and decisions that are hard for consumers to question. That mix is why regulators have focused on governance, testing and human oversight rather than banning the tools.

Key takeaways

  • Car insurers use AI in pricing, underwriting, telematics, claims estimates, customer service and fraud detection.
  • The NAIC adopted a model AI bulletin in December 2023; 25 jurisdictions had adopted it as of April 1, 2026, and four states have their own AI rules.
  • Insurers remain responsible for complying with insurance law when they use AI, including rules against unfair discrimination.
  • Treat an app's photo estimate as a starting point; get a shop estimate and ask for a human review if it looks low.
  • If a consumer report affects your price, you are entitled to a notice and can dispute errors in the report.

Terms used in this guide

Rating factors The characteristics insurers use to price a policy, such as driving record, location, vehicle, mileage and, where allowed, age and credit. Underwriting The process an insurer uses to decide whether to accept an applicant, on what terms, and in which pricing tier or company. Telematics Technology that records how, when and how much a car is driven, through an app, plug-in device or the car itself, so insurers can price on behavior. Claims adjuster The insurance company employee or contractor who investigates a claim, decides what the policy covers and sets how much is paid. Special investigations unit (SIU) The team inside an insurance company that investigates suspected fraud in claims and applications and refers cases to authorities. Credit-based insurance score A score built from parts of your credit report to predict how likely you are to file insurance claims, used in pricing where state law allows.

Related states

Colorado Colorado requires 25/50/15 liability; $5,000 MedPay and UM/UIM come with every policy unless rejected in writing. At-fault state with a 50% fault bar. New York New York is a no-fault state requiring 25/50/10 liability (50/100 for death), $50,000 of PIP and 25/50 uninsured motorist coverage. California California's minimum rose to 30/60/15 on January 1, 2025 and is set to reach 50/100/25 in 2035. It is an at-fault state with pure comparative negligence. Texas Texas is an at-fault state requiring 30/60/25 liability. PIP and uninsured motorist coverage come with every policy unless you reject them in writing.

Latest on this topic