Artificial intelligence now sets car insurance prices for most US drivers, and 88% of auto insurers already use it or plan to, according to a National Association of Insurance Commissioners (NAIC) survey. AI car insurance underwriting swaps broad actuarial tables for machine-learning models that score you on more than 100 data points, from your braking habits to your block-level ZIP code. By 2030, McKinsey projects more than 90% of personal auto pricing and underwriting will run fully automated.
AI car insurance underwriting uses machine-learning models to score your claim risk from 100+ data points, then prices your policy individually instead of by broad demographic group. About 88% of US auto insurers already use or plan to use AI, per the NAIC, and the technology cuts underwriting time by roughly 70%.
- AI scores you on 100+ data points, including telematics, credit, and ZIP code, versus the 10 to 15 variables traditional underwriting used
- Colorado became the first state to require insurers to test every pricing algorithm for bias, with auto compliance reports due July 1, 2026
- Fewer than 31% of telematics-enrolled drivers actually see a discount, per the Consumer Federation of America, while 24% get a rate increase
- Request a written explanation and human review whenever an AI-driven model raises your premium or denies a claim
What AI Underwriting Actually Means
Underwriting decides whether an insurer will cover you and at what price. Traditional underwriting sorts drivers into actuarial buckets using roughly 10 to 15 variables: age, ZIP code, driving record, vehicle, and credit in the 47 states that allow credit-based scoring. AI underwriting keeps those inputs and stacks hundreds more on top, then feeds everything through models that predict your individual probability of filing a claim.
Speed is the most visible change. AI-powered underwriting systems process applications about 70% faster than human reviewers, per industry data cited by Bankrate, and instant-quote engines now bind a policy in seconds. GEICO and Progressive both run behavioral and telematics data through pricing engines that can adjust continuously, not only once a year at renewal.
Granular is not the same as fair, though. A model that reads 100 signals can still bake in old patterns, which is why the way rating factors translate into your premium matters more than ever.
| Element | Traditional Underwriting | AI-Driven Underwriting |
|---|---|---|
| Data points used | 10 to 15 (age, ZIP, record, vehicle, credit) | 100+ (plus telematics, behavioral, external data) |
| Speed to quote | Hours to days | Seconds (about 70% faster) |
| Pricing basis | Broad actuarial groups | Individual claim-probability scores |
| Transparency | Filed rate tables, mostly explainable | Black-box models, harder to audit |
| How often it updates | Annually at renewal | Continuous or dynamic |
Source: NAIC AI survey data, Bankrate (2025), and McKinsey insurance research. Speed and automation estimates reflect personal auto and small-business lines.
The Data AI Uses to Set Your Rate
The shift from 15 inputs to 100-plus is the whole story. Where a human underwriter once read your application, an AI model now ingests streams of behavioral and external data, much of it collected after you buy the policy.
- Telematics and driving behavior: hard braking, rapid acceleration, mileage, time of day, and phone handling, which insurers convert directly into premium changes. See how that conversion works in our breakdown of driving behavior analytics.
- Credit-based insurance scores feed nearly every model, and a poor score can raise rates as much as 76% in some states.
- Location data drilled down to the ZIP code and sometimes the city block, capturing local theft, accident, and weather risk.
- Vehicle records covering repair cost, theft rate, and whether your car has driver-assistance features that lower claim severity.
- External datasets that range from aerial imagery to public records to purchased third-party data, the same category regulators flagged for review in 2026.
Telematics participation has scaled fast, with more than 21 million US policyholders now enrolled in usage-based programs. Compare the major telematics discount programs before you opt in, because the data you share rarely deletes.
Driving data has real cash value, and selling it without clear consent is now a legal liability. General Motors agreed to a $12.75 million settlement after driver data flowed to insurance brokers. Ask any insurer exactly what behavioral data it collects and who it shares the data with.
How AI Changes What You Pay
Safe drivers can come out ahead. Progressive reports average renewal savings of $322 a year, about 19% of a typical policy, for drivers who score well on its Snapshot program. That is real money: $322 saved works out to about $27 a month back in your pocket.
The catch is that most enrolled drivers never see that discount. The Consumer Federation of America found fewer than 31% of telematics participants got a price cut, while 24% saw a rate increase and 45% saw no change at all.
| Telematics Outcome | Share of Enrolled Drivers | Typical Annual Impact |
|---|---|---|
| Premium decreased | Under 31% | -$120 to -$332 |
| No change | 45% | $0 |
| Premium increased | 24% | Higher rates |
Source: Consumer Federation of America analysis of telematics outcomes, with dollar ranges drawn from Consumer Reports ($120 median) and Policygenius ($332 average) savings data.
The deeper worry sits in the proxy data. A model never asks your race, yet ZIP code, credit, occupation, and education all correlate with it, so bias slips in through the side door. The Consumer Federation of America found drivers in predominantly Black communities pay 71% more for auto coverage, which on a $1,700 policy adds roughly $1,200 a year, close to $100 every month.
| Where the Gap Shows Up | Extra Cost vs. Comparable Drivers | Monthly Impact |
|---|---|---|
| Predominantly Black communities (US) | +71% (about $1,200/yr) | ~$100/mo |
| Non-white ZIP codes (New York) | +$1,728/yr | ~$144/mo |
Source: Consumer Federation of America and New York rate analysis cited by MoneyGeek (2026). Figures compare drivers with similar records and vehicles across different communities.
An algorithm that reads 100 signals can still discriminate without ever naming race. ZIP code and credit score do the work that the law forbids asking directly, which is exactly why state regulators stepped in.
Which States Are Pushing Back
Regulators are no longer watching from the sidelines. The NAIC adopted a Model Bulletin on the Use of AI Systems by Insurers in December 2023, and 24 states plus the District of Columbia have since put a version in place. A new AI Systems Evaluation Tool launched as a 12-state pilot in March 2026 and is slated for nationwide use by November 2026.
State insurance commissioners opened a formal review of how carriers use AI in pricing, a move detailed in our report on the regulatory review of AI auto pricing. Colorado went furthest, becoming the only state that requires insurers to inventory and bias-test every pricing algorithm.
| State or Body | Rule | What It Requires | Status |
|---|---|---|---|
| Colorado | SB 21-169, Reg 10-1-1 | Bias-test every pricing algorithm; annual reports | Auto reports due July 1, 2026 |
| New York | DFS Circular Letter 2024-7 | Bias testing plus explainability for AI and external data | Effective 2024 |
| NAIC (24 states + DC) | Model Bulletin on AI | Principle-based governance framework | Adopted since Dec 2023 |
| 12 pilot states | AI Systems Evaluation Tool | Insurers submit AI systems for regulator review | Pilot March 2026; national Nov 2026 |
| California | SB 1120 | Human review required before AI-only coverage denial | Effective Jan 2025 (health focus) |
Source: NAIC AI issue brief (March 2026), Colorado Division of Insurance, and MoneyGeek regulatory tracking. Auto-relevant rules vary by state.
Carrier AI strategy is moving fast. State Farm announced an OpenAI partnership and AI claims tools in 2026, a sign that the biggest insurers are betting heavily on automated pricing and claims handling.
How to Protect Yourself From Unfair AI Pricing
You cannot turn off the algorithms, but you can shop around them and demand transparency. These four steps put the leverage back on your side.
Compare quotes from 3 to 5 carriers
Every insurer trains its model differently, so the same driver can see swings of hundreds of dollars. Comparison shopping is still the single biggest lever on your rate.
Ask whether AI sets your price
In Colorado and New York, insurers must disclose AI use on request. Ask which data fed your quote, then correct anything wrong, since a single bad data point can move your premium.
Treat telematics as optional, not automatic
With 24% of participants seeing increases per the CFA, weigh the odds first. Review how usage-based insurance scores your trips before you plug in a device.
Demand human review and file a complaint
Request a written explanation for any rate hike or denial, ask for a human to re-review, and report unfair pricing to your state insurance commissioner, which creates an enforcement record.
Re-shop at every renewal, not just when your rate jumps. Because AI models reprice continuously, the carrier that was cheapest last year may rank third this year, and switching often saves $400 or more annually.
Frequently Asked Questions
It depends on your profile. Safe drivers can save, with Progressive reporting average Snapshot savings of $322 a year, but the Consumer Federation of America found fewer than 31% of telematics users get a discount and 24% see an increase. Drivers in some communities pay up to 71% more due to proxy data.
AI models use 100+ data points: telematics (braking, speed, mileage, time of day, phone use), credit-based insurance scores, ZIP-code and block-level location risk, vehicle repair and theft data, and external datasets like aerial imagery and public records.
You cannot fully opt out, since nearly 9 in 10 auto insurers use AI somewhere in pricing per the NAIC. You can decline telematics tracking, which is voluntary, and shop carriers that weight your individual record more heavily than behavioral data.
Yes, but it is increasingly regulated. Colorado requires insurers to bias-test every pricing algorithm, New York mandates explainability under DFS Circular Letter 2024-7, and 24 states plus DC have adopted the NAIC AI model bulletin.
Ask directly. In Colorado and New York, carriers must disclose AI use and the data behind your quote on request. You can also request a written explanation of any premium change and ask for human review of an AI-driven decision.
- NAIC: Artificial Intelligence in Insurance
- NAIC: Artificial Intelligence and State Insurance Regulation (March 2026)
- Bankrate: How Your Insurance Company Is Using AI
- MoneyGeek: How AI Is Changing Insurance in 2026
- McKinsey: The Future of AI in the Insurance Industry
- Consumer Federation of America: Auto Insurance Research
- Deloitte: Colorado Division of Insurance AI Regulation
- U.S. News: Will AI Raise Your Insurance Premiums? New Rules That Protect You