AUTOBODY DIRECTORY
Home › Guides › AI Is Grading Your Auto Claim, and Almost No Law Requires a Human to Check It

AI Is Grading Your Auto Claim, and Almost No Law Requires a Human to Check It

Short answer: software already decides more of your auto claim than most people think. The photo estimate, the triage that routes your file, the total-loss value and even the letters are machine work at a growing number of carriers. That is not automatically bad. The problem is when an automated system becomes the reason a claim is cut or denied with no real human judgment behind it. The lawsuits over exactly that, and the one law that forces a human to review it, are in health insurance. Auto's own version of that law died in a Florida committee this year. So for now, the only guarantee that a licensed human stands behind your car's number is the old adjuster-licensing rules, not any AI-specific requirement. None of this is legal advice. Here is what is real and what you can do.

By Anthony Braswell for Quorum Industries LLC, The Autobody Directory · Updated 2026-08-16 · How this was written, and what the machine may not do

How this page was produced: Researched, drafted and checked with AI assistance under human direction, and signed off by the named author. How this site is written

How much of your claim is already automated

This is not coming, it is here, and the carriers say so themselves.

Allstate's photo-estimate app handles roughly half of its driveable-vehicle claims and cut the estimating cycle from five to seven days down to under a day. On its third-quarter 2025 earnings call, Allstate said that in claims, all adjuster emails are now generated or reviewed by AI, and it drafts on the order of 50,000 claim communications a day. Progressive told investors it deployed photo-estimating AI in place of 200 staff it would otherwise have hired. State Farm holds hundreds of AI patents covering claims triage, fraud flags, and estimating damage severity from photos.

Across the industry, a national insurance-regulator survey found roughly 88 percent of auto insurers use or plan to use AI, and that nearly a third of insurers do not regularly test their models for bias. Speed is the selling point. The open question is what happens when the fast answer is also the wrong one.

The lawsuits, and why they are all in health insurance

The cases that put AI claim denials in the headlines are health insurers. Cigna was sued over a tool that, plaintiffs say, reviewed each request in about 1.2 seconds and denied hundreds of thousands of claims in a matter of months, short-circuiting the physician review its state required. UnitedHealth and Humana were sued over a model that plaintiffs say carried a 90 percent error rate on appeal, and whose rollout tracked a jump in denials. Those cases are active.

They matter to a car owner for one reason: the protections that grew out of them are health-only.

The law that requires a human, and why it may not cover your car

California's SB 1120, in force in 2025, bars a health insurer from denying coverage based solely on an algorithm and requires a qualified human review. It does not reach auto. The national association of insurance regulators issued an AI governance bulletin in December 2023, adopted by two dozen-plus states, but it is guidance about how insurers should govern and document their systems, not a hard rule that a human must review each auto decision.

Auto's own human-review bill, and how close it came

Florida tried to close that gap. HB 527 and its companion SB 202, in the 2026 session, would have barred AI from being the sole basis to deny or reduce any insurance claim, auto included, and required a qualified human professional to independently analyze the claim and check whatever the algorithm produced before a denial. It passed the Florida House 108 to 0. Then it died in Rules on March 13, 2026, before becoming law. Our own read of that margin is that a bill with no recorded opposition in one chamber tends to return in the next session, though nothing obliges it to. If you have heard that a licensed human is supposed to sign off on a denial, this is almost certainly what you heard: it is real, it was one step from passing, and it is not law yet.

Texas told its insurers a person has to agree with the machine

The gap Florida did not close, Texas came at sideways, and the difference between the two is the point.

On 12 June 2026 the Texas Department of Insurance issued Commissioner's Bulletin B-0003-26, addressed to "All regulated entities and their agents and representatives." Its opening line is the whole idea: decisions or actions impacting consumers "that are made or supported by advanced analytical and computational technologies, including artificial intelligence (AI), must comply with all applicable insurance laws and regulations."

Then it does something more useful than stating a principle. It lists the statutes it means, and two of them are exactly the ones this page has been pointing at. Texas Insurance Code Chapter 542, unfair claims settlement practices, which the bulletin describes as setting "standards for the investigation and disposition of claims arising under policies or certificates of insurance issued to residents of Texas." And Chapter 4101, insurance adjusters, which the bulletin describes in these words: it "requires that certain acts may be taken only by a person who holds a license to act as an adjuster. Official acts of an adjuster include investigating or adjusting losses and supervising the handling of claims."

Read those two together and the argument in the next section stops being an inference we are drawing. A state regulator has put adjuster licensing on its own list of laws that an AI-assisted claim decision has to satisfy.

The bulletin then states an expectation about people, in one sentence: "If a regulated entity uses AI to make a consequential decision, TDI expects a person to review and agree with all decisions before action is taken." Note the verb pair. Not review. Review and agree. Agreement is a position a person has taken, and a position can be examined.

It also states the standard the output has to meet no matter how it was produced. Decisions must not be "inaccurate, arbitrary, capricious, or unfairly discriminatory", and "compliance with such standards is required, regardless of the tools and methods used in decision-making."

Now the limits, because they are large. A Commissioner's Bulletin is a department's statement of its expectations. It is not a statute and it is not a rule, and this one says as much about itself: "The goal of this bulletin is not to prescribe specific practices or documentation requirements." "Expects a person to review and agree" is an expectation, not a codified duty, the bulletin does not define "consequential", and none of it reaches outside Texas. Nothing here changes the headline on this page.

One more thing the bulletin makes visible by contrast. The single flat prohibition on AI in its whole list sits somewhere else entirely: Chapter 4201, utilization review agents, which the bulletin says "prohibits using AI to make an adverse determination." That is health care utilization review. It is not auto claims. Texas does have a statute telling insurers a machine may not make the adverse call, and it does not cover your car, which is the same pattern the rest of this page describes.

So what genuinely changed on 12 June 2026 is smaller than the headlines around it and more usable than nothing. The enforceable law is what it already was. What is new is a written record of what the regulator will ask for. TDI says it "will monitor the use of AI through examinations and product filings", that it "welcomes any complaints from consumers regarding the use of AI and will investigate accordingly", and that all regulated entities using AI "should be able to furnish procedures and protections upon request by TDI." If you are in Texas and you believe software set your number, that last clause is the sentence to put in the complaint.

So what actually protects your auto claim today

Not an AI statute. The older rule that adjusting and appraising a claim is a licensed activity. In Florida, a licensed all-lines adjuster (the 6-20 license) is who may legally settle a claim and set the amount payable. South Carolina and Massachusetts license motor vehicle damage appraisers specifically.

Here is the point, and its limit. In a state where adjusting or appraising a claim is a licensed activity, an algorithm's output does not itself become the licensed professional. A licensed human has to adopt that number, and a person who adopts it is accountable for it in a way software is not. What that duty covers, and how far it reaches into an automated workflow, varies by state and by which activity is being performed. So the question worth asking is not whether AI touched your claim. It is which licensed person is standing behind the number, and on what basis.

What to do

  1. Ask who, or what, decided. In writing, ask whether a licensed adjuster reviewed your file and the estimate, or whether the number and any denial were automated. Put the answer in your records.
  1. Make a machine-made denial name a human. If a denial or a low number looks automated, ask for the qualified person's basis for it, not just the output.
  1. Dispute the specifics and use the clause. Challenge the estimate or valuation line by line (see the guides below), and for amount disputes invoke your policy's appraisal clause, which puts real appraisers on both sides.
  1. File a complaint with your state insurance department. It is free, it is leverage, and in states weighing an AI-review bill it is also a data point lawmakers count. Find your state's regulator and its complaint page.

A regulator says the quiet part out loud

Virginia's Bureau of Insurance puts a notice on the page where you file a complaint. It warns that AI tools "can often provide incorrect information about insurance," that they may misinterpret policy documents and state insurance laws, and it asks consumers to verify anything an AI tells them against other sources. You can read it on the Bureau's file an insurance complaint page.

Sit that next to the rest of this guide. The insurer is running software on your claim, and the regulator you would complain to is telling you not to trust software about your claim either. Those are not contradictory, and the resolution is the same in both directions: get the specific thing in writing, from the party who is accountable for it. An AI answer about your policy is a starting point for a question, never the answer. So is this guide. The difference is that this one shows you where it got everything.

What this page is not saying

We are not saying an insurer used AI to decide your claim. This page reports what carriers have said about their own automation, what regulators have written, and what the law does and does not require. What happened in any particular file is not something we know or claim to know.

We are not saying any of the health-insurance cases decided anything about auto. The lawsuits and the human-review law described here are health-insurance matters. They are on this page because auto has no equivalent, and that absence is the finding.

We are not claiming the Texas bulletin is a statute. It is the department telling its insurers what it expects, and we say so where it appears. A bulletin can be withdrawn, and it is not enforced the way a statute is.

We do not claim the survey figures describe your carrier. An industry-wide percentage says nothing about which side of it any one company sits.

This is not legal advice.

Where this fits

For the disputes this most often shows up in, see total loss: why the first check is low, when insurance won't pay for a proper repair, and how insurance claims work. For the paperwork that proves what your car actually needed, see scan reports and what they prove.

Related

Sources

The internal guides linked above carry their own citations. For the outside facts, here is the trail. Where a link appears below, we read that page on the issuing body's own site before linking it. Where no link appears, the fact came from company disclosures or from industry and legal coverage, and we say so rather than dress it up as a primary citation.

General consumer information: not legal, insurance, or financial advice. Requirements, coverage, and practices vary by state, policy, and manufacturer.

Run a body shop? Your shop likely already has a page here, built from public records. Check it and claim it free: verifying only ever adds.
What does claiming add? It's free ›

Where this fits

Each link says what it is for. We add one only when a reader on this page has a real reason to need that page next.