When an automated fraud flag delays your claim

A claim that goes quiet after a routine filing has often been scored by software and routed to a person who is checking it. The flag is a reason to look closer, not a finding that you did anything wrong, and the insurer is allowed to take longer than usual when it has a documented reason to suspect fraud. Your job is to keep the file complete, answer reasonable requests in writing, and use the deadlines that still apply to press for a decision.

Most drivers who hit this learn about it only by its effects: an adjuster who stops returning calls, a request for documents you already sent, a letter saying "additional time is needed" with no reason you can parse. This article explains what happens behind the flag, what the insurer may ask, what the regulations say about timing, and how to protect your position without turning a delay into a fight.

What an automated flag is

Insurers run claims through scoring models and data checks, usually within hours of the first notice of loss. A model compares the new claim with patterns drawn from past claims, including those later found to be fraudulent, and gives it a score. A claim over a threshold goes to a person, often someone in a special investigations unit (SIU), for review. How a database such as ISO ClaimSearch links claims is covered in ISO ClaimSearch and how fraud analytics flag claims; this article is about what the flag does to your timeline.

Three points frame everything that follows.

First, a flag is a triage decision. The model is deciding where a human should look, not who is lying. The National Insurance Crime Bureau makes the same point about its own indicator lists, which it says only "indicate" the potential of fraud and "are not actual evidence" (NICB's scene indicators for law enforcement).

Second, the regulator's rules about claims handling still apply. The NAIC's 2023 model bulletin on insurers' use of AI systems lists fraud detection among the uses of AI in insurance and says actions taken by insurers must not violate the state's unfair claims settlement practices law "regardless of the methods the Insurer used." A model is a method, not an exemption. The bulletin also tells insurers that the extent of human involvement in a final decision is one factor in how much oversight a tool needs, and that regulators may request documents about it during an examination. States adopt the bulletin on their own, so ask your department whether yours has.

Third, most flagged claims are real losses. The NAIC describes soft fraud, which is padding a legitimate claim or lying on an application, as more common than staged losses. That means reviewers spend much of their time sorting honest claims with odd features from claims with real problems.

What models weigh, and what the innocent explanation might be

No insurer publishes its scoring formula, and anyone who tells you the exact weights is guessing. What can be said is which kinds of signals fraud reviewers and investigators generally treat as worth a second look, and why an honest claim can show the same signal. NICB's indicator list is one public source for the pattern: vague answers about the facts of a theft, a vehicle that had been advertised for sale, an out-of-state phone number, injuries that are subjective rather than visible. Others come from how insurers describe their own review.

The last column is the heart of the problem. Every one of these signals has a lawful explanation, which is why a flag should lead to a question and not a denial.

How false positives happen

False positives come from a few predictable places.

• Thin context. The model sees dates and amounts, not your life. A recent policy change looks the same whether it came from a scheme or from buying a new car.
• Base rates. If one claim in a hundred is fraudulent, even an accurate model flags many honest claims, because honest claims vastly outnumber fraudulent ones. This is arithmetic, not malice.
• Data errors. A mismatched name, a transposed VIN digit or a reused repair-shop address can produce a "link" that does not exist. Errors in the claims-history databases you can pull yourself are covered in Pulling your own records: CLUE, MVR and claim file.
• Third-party data. A scoring tool may rely on data from vendors, and the NAIC bulletin asks insurers to have due diligence for third-party data and AI systems.
• Pattern drift. Models trained on past schemes can mistake a new, legitimate behavior, such as a rise in glass claims after a hailstorm, for an old one.

A human reviewer is supposed to catch these. When the review is thin or the reviewer is overloaded, the practical effect on you is delay.

What happens after a referral

Once a claim reaches the SIU, expect some or all of the following, depending on the type of loss.

1. A call from a different person. An investigator may introduce themselves and ask for a recorded statement. recorded statements: when to give one and when to wait explains how to prepare for one and when to ask to reschedule so you can gather documents first.
2. Document requests. Typical requests: purchase paperwork, repair records, photos, phone records for the time of the loss, medical records for injury claims, and proof of how the car was used.
3. Inspection. The car, the scene or the damaged parts may be examined, and for a theft, keys and the car's history may be requested.
4. Third-party checks. Interviews with the other driver, witnesses or repair shops.
5. An examination under oath. In a serious case the insurer may require a formal sworn interview under the policy's cooperation clause. The mechanics are in the cooperation clause: records, interviews and exams under oath.

The insurer can ask for information reasonably related to the claim. California's regulation puts the limit plainly: the insurer must conduct a "thorough, fair and objective investigation" and "shall not persist in seeking information not reasonably required for or material to the resolution of a claim dispute" (10 CCR 2695.7(d)). Other states have similar language in their own rules, but the wording varies, so check yours. A request for ten years of bank statements on a $1,800 glass claim is the kind of request worth questioning in writing.

Also expect secrecy. California's regulation says nothing in its denial or delay-notice provisions requires an insurer to disclose information that could reasonably be expected to alert a claimant that the claim is being investigated as a suspected fraudulent one (2695.7(b)(2) and (c)(2)). So "we need more time" can be the entire explanation you get. That is lawful in that state, and it is why your best evidence of a flag is the pattern of delay.

Which clocks still run

Claims handling rules set deadlines for acknowledging, investigating and deciding a claim, and fraud suspicion changes how some of them apply.

The NAIC's Unfair Property/Casualty Claims Settlement Practices Model Regulation (#902), in its 1997 text, sets this model timeline:

• Acknowledge a claim within 15 days of notification.
• Tell a first-party claimant whether the claim is accepted or denied within 21 days after receiving properly executed proofs of loss.
• If more time is needed, say so within those 21 days, give the reasons, and send a letter explaining why more time is needed 45 days after that notice and every 45 days after.
• Pay within 30 days of affirming liability, when the amount is determined and not in dispute.
• Put a denial in writing and cite the policy provision it relies on.

Then comes the exception that matters here. Where the insurer has "a reasonable basis supported by specific information available for review by the insurance regulatory authority" for suspecting that the claimant fraudulently caused or contributed to the loss, it is relieved of the 21-day and 45-day requirements, but must still advise the claimant of acceptance or denial "within a reasonable time for full investigation" after proof of loss. The key words are "supported by specific information available for review." A bare model score is not a documented reason; an investigator's file note is closer.

States adopt their own versions, and the numbers differ. California's rule, for example, requires an insurer to accept or deny within 40 calendar days of proof of claim, to send written notice of the need for more time within that period with the information still required, and to repeat the notice every 30 calendar days until a decision is made (10 CCR 2695.7(b) and (c)(1)). It also requires the denial letter to tell the claimant that the California Department of Insurance can review it. Whatever your state's numbers, the model and the California rule agree on the ideas you can use: a decision within a stated time, periodic written notice of why more time is needed, and written denials that cite the policy language. Our guide to unfair claims practices acts: the deadlines insurers must meet compares the deadlines across states.

One more clock deserves attention. The model regulation requires an insurer to give an unrepresented claimant written notice of a statute-of-limitations deadline at least 30 days before it expires for first-party claimants and 60 days for third-party claimants. Time limits in your policy, such as a suit-limitation clause, still run during an investigation. Do not let a long delay run out a deadline without checking it, and ask the department or an attorney about the dates that apply to you.

A worked example

Say your car is stolen from a driveway and you file on a Monday. Eight days earlier you had changed your policy to add comprehensive coverage because you had bought the car with cash from a private seller. Both facts raise a score. These details are invented for the exercise.

1. Day 0: you report the theft to police and the insurer. You send the bill of sale, the title, photos of the car and both keys.
2. Day 3: an SIU investigator calls and asks for a recorded statement the next day. You say you will give one, but ask for a few days to gather paperwork, and you confirm by email.
3. Day 10: you give the statement. You send, in one email, the purchase records, the seller's contact details, your bank record of the payment, service receipts and your phone's location history for the night, and ask for confirmation of receipt.
4. Day 24: you receive a letter that more time is needed. You note the date. Under a rule that requires a decision within a set period, you calendar the next date.
5. Day 45: nothing. You email the adjuster and the investigator asking what specific additional information is needed and referencing the date of your last submission.
6. Day 60: with no answer, you file a complaint with your state insurance department and attach the timeline.

Nothing here accuses anyone. It builds a record that shows you were responsive and that any delay was theirs.

How to protect your position

• Answer in writing. Confirm calls by email: who, when, what was asked, what you agreed to send. The keeping a claim diary to protect yourself article lays out a format.
• Send complete packets. One organized submission with an index beats five partial ones. Keep copies and proof of delivery.
• Ask what is outstanding. "Please list every item you still need to decide this claim" turns an open-ended delay into a checklist.
• Don't guess. If you do not remember a detail, say so; do not fill the gap. Inconsistencies are what investigators look for.
• Know your policy's conditions. The cooperation clause is real. Declining reasonable requests can give an insurer a reason to deny, which is different from asking to schedule or narrow a request.
• Use the department. When a deadline passes or a request looks excessive, your state insurance department takes complaints. The NAIC's guidance on filing a complaint suggests having your name, the type of insurance, the reason for the complaint, supporting documents, correspondence and a log of calls ready.
• Get advice if the stakes are high. If you are asked for an examination under oath, are told your claim may be denied for misrepresentation, or face a large loss, a licensed attorney can advise on your state's rules in a way this article cannot.

Fraud-reporting immunity: what it covers

Insurers often say they cannot discuss details of a fraud referral. A related point is that many states protect people who report suspected fraud from civil suits. Texas is an example. Under Insurance Code section 701.052, a person is not liable in a civil action, including libel or slander, for furnishing information about suspected, anticipated or completed fraudulent insurance acts to specified recipients such as the insurance department, law enforcement, the NAIC or an insurer's special investigative unit. The protection does not apply to someone acting with "malice, fraudulent intent, or bad faith," and a protected person who wins a suit over such a disclosure is entitled to attorney's fees and costs. Section 701.051 separately requires a person who reasonably suspects fraud to report in writing within 30 days.

What this means for you: an insurer that refers your claim in good faith is generally protected from a suit over the referral itself. What it does not do is excuse a bad-faith delay, an unreasonable demand, or a baseless denial. Those are handled by claims-handling rules, department complaints and, in some states, lawsuits. Our article on mandatory fraud reporting and immunity laws for insurers compares the state approaches; check your own state's law, since the Texas text above applies only in Texas.

Common questions

Will the insurer tell me my claim was flagged for fraud?

Often not. California's regulation, for one, expressly does not require an insurer to disclose information that could alert a claimant to a fraud investigation. Judge the situation by behavior: a referral to the SIU, requests for recorded statements or sworn examinations, and repeated delay notices.

Can the insurer refuse to pay while it investigates?

It can take reasonable time to investigate. Under the model regulation it can take longer than usual if it has a documented basis for suspecting fraud, but it must decide within a reasonable time and cannot hold the claim indefinitely. If the amount of some part of the claim is not in dispute, ask whether that part can be paid.

Does a fraud flag go on my record?

A claim, paid or not, may be recorded in industry claims databases. An internal flag or SIU referral is not a conviction, and there is no public source showing that a flag by itself becomes a permanent mark on a driver. Request your own records to see what is held, as explained in our guide to Your CLUE report and motor vehicle record explained.

Should I get a lawyer right away?

Not necessarily for a routine delay. Consider one if you receive a request for an examination under oath, a written accusation of misrepresentation, a denial based on fraud or a coverage dispute involving serious money. A department complaint is usually a lower-cost first step.

What if I made an honest mistake on the application or the claim?

Tell the truth and correct it in writing. claim padding vs fraud: where the legal line falls explains where honest error ends and a misrepresentation begins, and how it can affect coverage.

Key takeaways

• An automated flag is a triage step that sends a claim to a human reviewer; it is not evidence of fraud, and every common red flag has an innocent explanation.
• The NAIC's model claims regulation relaxes its 21-day and 45-day clocks only where the insurer has a reasonable basis, supported by specific information that a regulator can review, to suspect fraud, and it still requires a decision in a reasonable time.
• Some states, California among them, do not require an insurer to tell you a claim is under fraud investigation, so judge by the pattern of delays and requests.
• Keep a written record, send complete document packets, ask what is outstanding, and answer reasonable requests; unreasonable ones can be questioned in writing.
• If deadlines pass without a decision, a complaint to your state insurance department is the next step, and a licensed attorney is the place to go for sworn examinations or accusations of misrepresentation.