Why You Should Never Use ChatGPT to Settle Your Own Car Accident Claim in California

Maria spent four hours one Saturday night doing what she thought was the smart, modern thing. She had been rear-ended on the 405 three weeks earlier, her neck and lower back were still in constant pain, and she was already drowning in medical bills. Hiring a lawyer felt expensive and intimidating, so she did what millions of Americans now do when they face a hard problem: she opened ChatGPT.

She typed in the police report number, her injuries, the name of the at-fault driver’s insurance company, and asked for a demand letter for $85,000. Within seconds, the AI produced a polished, professional-looking document. It cited California law. It used phrases like “general damages” and “pain and suffering.” Maria signed it, attached her medical bills, and emailed it to the adjuster.

Eleven days later, the response came back: a settlement offer of $4,200, on a take-it-or-leave-it basis, with a thirty-day deadline. The adjuster’s email was friendly. It was also a trap.

Maria’s case, properly investigated, properly documented, and properly leveraged, was worth somewhere between $90,000 and $140,000. By the time she walked into my office, the statute of limitations clock had been ticking for nearly a year, key evidence had vanished, and she had given a recorded statement to the insurance company that we now had to spend months neutralizing.

Maria is not alone. Since the public release of ChatGPT, my firm has seen a sharp increase in injured Californians attempting to handle their own car accident claims using AI tools. The pattern is almost always the same. They get a result that looks professional. Then they lose tens of thousands of dollars they will never recover.

I have spent thirty years representing injured people in Los Angeles and across California. I have tried cases to verdict, taken hundreds of depositions, and negotiated thousands of settlements with every major insurance carrier in this state. What follows is an honest, practical explanation of why AI cannot do what an experienced personal injury lawyer does — and what is really at stake when you try.

The Misconception About AI in Personal Injury Cases

Let’s start by being fair to the technology. Modern AI tools are genuinely impressive. They can draft letters that read smoothly. They can summarize legal concepts at a high level. They can organize information, generate timelines, and produce documents in seconds that would take a paralegal hours to type.

All of that is true. None of it matters when you are trying to settle a real injury claim.

Here is the misconception. People assume that because a demand letter is a document, and AI is good at producing documents, AI must be good at producing demand letters. But a demand letter is not really a document. It is the visible end product of a months-long process of investigation, evidence gathering, medical documentation, damages calculation, coverage analysis, and strategic positioning. The letter itself is maybe two percent of the work.

AI can write the two percent. It cannot do the other ninety-eight percent.

More importantly, AI lacks the four things that actually move insurance companies:

• Legal judgment — knowing what arguments will actually work against this adjuster, this carrier, in this venue, on these facts.
• Strategic timing — understanding when to push, when to wait, when to file suit, and when to take a deposition.
• Negotiation experience — recognizing the patterns adjusters use, knowing when an offer is final and when it is a feeler.
• Litigation power — the credible, demonstrated ability to file a lawsuit and try the case in front of a jury if the offer is unfair.

ChatGPT has none of these. It cannot pick up the phone, file a complaint, take a deposition, or walk into a courtroom. The insurance industry knows this, which is why AI demand letters do not scare them.

Everything That Must Happen BEFORE a Demand Letter Goes Out

This is the part that AI users almost never understand. By the time a demand letter is written, the case is largely already won or lost. The work that determines settlement value happens in the months before that letter is drafted — and almost none of it can be performed by an AI.

1. Liability Investigation

Establishing fault is rarely as simple as “the other driver hit me.” California is a comparative fault state, which means the insurance company will work hard to assign you a percentage of blame — even when you did nothing wrong. Every percentage point they pin on you reduces your recovery by that exact amount.

A proper liability investigation includes:

• Obtaining and analyzing the official traffic collision report (and supplemental reports)
• Locating and interviewing independent witnesses before their memories fade
• Identifying nearby surveillance cameras (gas stations, ATMs, businesses, doorbell cameras) and sending preservation letters before footage is overwritten — typically within 30 days
• Requesting and decoding the at-fault vehicle’s event data recorder (EDR), commonly called the “black box”
• Retaining an accident reconstruction expert when speed, angle, or sequence of impact is in dispute
• Pulling 911 audio, dispatch logs, and CHP MAIT (Multidisciplinary Accident Investigation Team) reports for serious crashes

ChatGPT cannot send a preservation letter. It cannot interview a witness. It cannot get a black box downloaded before the totaled vehicle is sold for salvage and the data is lost forever.

2. Evidence Preservation and Spoliation

In California, when evidence relevant to a potential claim is destroyed by a party who knew or should have known of the claim, courts can impose spoliation sanctions — including jury instructions that the destroyed evidence would have been unfavorable to that party. But spoliation sanctions only help you if you sent a proper preservation letter, in writing, to the right parties, before the evidence disappeared.

Dashcam footage gets overwritten on a loop. Surveillance video is typically purged within 14–30 days. Commercial trucks have ELD (electronic logging device) data that can be erased after eight days. Smartphones get wiped. Vehicles get repaired or junked.

An AI cannot identify what evidence exists, who controls it, or how to legally compel its preservation. By the time most self-represented claimants think to ask, the evidence is already gone.

3. Medical Documentation and Treatment Strategy

This is where AI users do the most damage to their own cases. Insurance adjusters do not value injuries based on how badly you hurt. They value injuries based on what is documented in your medical records — and how it is documented.

A few examples of what experienced personal injury counsel does that AI cannot:

• Coordinates referrals to the right specialists (orthopedists, neurologists, pain management, neuropsychologists for traumatic brain injuries)
• Identifies and addresses gaps in treatment that adjusters use to argue you weren’t really hurt
• Ensures providers properly document causation — connecting your injuries to the crash in the chart, not just listing symptoms
• Obtains diagnostic imaging (MRI, CT) when indicated, instead of relying solely on X-rays that may show “normal” findings even with serious soft-tissue damage
• Builds a future medical care projection through life care planners and treating physicians for serious or permanent injuries
• Manages medical liens so the lien holders (health insurers, Medicare, Medi-Cal, hospitals) don’t swallow your settlement

ChatGPT will tell you that pain and suffering is compensable. It will not tell you that your chiropractor’s note saying “patient reports neck pain” is worth roughly nothing to an adjuster compared to a properly documented MRI showing a C5–C6 disc protrusion with radiculopathy.

4. Damages Calculation

Calculating damages in a California personal injury case is far more complex than adding up your medical bills and multiplying by three. A real damages analysis includes:

• Past medical specials (billed amounts, paid amounts, and what’s recoverable under Howell v. Hamilton Meats)
• Future medical expenses, supported by expert opinion
• Past and future lost wages, including loss of overtime, bonuses, and benefits
• Loss of earning capacity for permanent injuries (often supported by a vocational expert and an economist)
• General damages — pain, suffering, mental anguish, loss of enjoyment of life — calibrated to recent jury verdicts in your venue
• Loss of consortium claims for spouses, where applicable
• Property damage and diminished value

AI tools routinely produce demand numbers that are either grossly inflated (which destroys credibility with the adjuster) or grossly understated (which costs you tens of thousands). Neither is a good outcome.

5. Insurance Coverage Analysis

This may be the single most overlooked area in self-represented claims. Most people assume there is one insurance policy: the at-fault driver’s liability coverage. In reality, a typical California auto case may involve multiple layers of coverage:

• The at-fault driver’s bodily injury liability policy
• An excess or umbrella policy carried by the at-fault driver
• If the at-fault driver was working — a commercial auto policy and possibly an employer’s general liability policy
• If a rideshare or delivery app was involved — Uber, Lyft, DoorDash, Amazon Flex policies (often $1 million during the active period)
• Your own uninsured/underinsured motorist (UM/UIM) coverage
• MedPay coverage on your own auto policy
• Health insurance, with subrogation rights to be negotiated

ChatGPT does not know what policies exist. It cannot send a demand for policy limits disclosure. It cannot evaluate whether a third-party defendant has assets beyond the policy. And it absolutely cannot identify when your own UM/UIM policy needs to be opened to fully compensate you for a serious injury.

Why AI-Generated Demand Letters Fail

Insurance adjusters handle thousands of claims a year. They can spot an AI-generated demand letter within the first paragraph. Here is what they see, and how it changes their valuation in real time:

Pattern Recognition by Adjusters

AI letters share telltale features: generic structure, formulaic recitation of “general damages” without specific facts, identical phrasing across cases, no reference to specific medical providers or diagnostic findings, no analysis of the police report, and a demand number that does not match the documentation. Major carriers — State Farm, GEICO, Allstate, Farmers, Mercury, Liberty Mutual, Progressive — have internal training and, increasingly, their own AI tools to flag these letters automatically.

Lack of Supporting Evidence

A persuasive demand package is not just a letter. It is a binder of evidence: the police report with a written analysis of liability, certified medical records, itemized medical bills, wage loss documentation from your employer, photographs of injuries and property damage, witness statements, expert reports, and case law citations relevant to your specific facts.

AI users almost never assemble this package, because they don’t know it’s expected. The adjuster reads the letter, sees no exhibits, and concludes — correctly — that there is no real case behind the demand.

The Unrepresented Claimant Signal

Insurance carriers maintain internal data on settlement outcomes. They know, with statistical certainty, that unrepresented claimants accept less. The Insurance Research Council, an industry-funded research organization, has repeatedly published findings that represented claimants recover dramatically more on average than unrepresented claimants — even after attorney fees.

When an adjuster receives an AI-generated demand letter from a self-represented claimant, the calculation is simple: this person has no lawyer, no investigator, no medical expert, no ability to file suit, and no leverage. The opening offer is set accordingly — often at fifteen to twenty cents on the dollar of true case value.

No Litigation Threat Means No Leverage

This is the single most important concept in injury claims, and the one AI users miss completely.

Insurance companies do not pay fair value because they read a persuasive letter. They pay fair value because they fear what happens if they don’t. That fear is the credible, documented threat of litigation.

A real litigation threat means:

• Filing a complaint in superior court before the statute of limitations runs
• Serving the defendant and putting them in default if they don’t answer
• Conducting written discovery — interrogatories, document requests, requests for admission
• Taking the deposition of the at-fault driver, the adjuster’s chosen experts, and any percipient witnesses
• Retaining your own qualified experts: biomechanical engineers, medical specialists, economists, accident reconstructionists
• Filing and arguing motions in limine to control the evidence at trial
• Trying the case to verdict in front of a jury

Each of these steps costs the insurance company money. Defense counsel rates in California typically run $250 to $500 per hour. Expert witnesses charge $500 to $2,000 per hour for deposition and trial testimony. A case that goes to trial can cost the insurer $75,000 to $250,000 in defense costs alone — before the verdict.

That cost is your leverage. The adjuster’s job is to resolve the claim for less than the cost of defending it. If they know you cannot file a lawsuit — because you have no attorney, you don’t know how, and the AI cannot do it for you — that leverage evaporates.

Common Mistakes People Make Using AI to Handle Their Own Claim

Over the last two years, I have personally handled cases that came to my office only after the client had tried to settle with AI. The same mistakes appear over and over.

1. Settling Too Early

Soft tissue injuries can take six to twelve months to fully manifest. A herniated disc may not be diagnosed until an MRI is finally ordered four months post-crash. Traumatic brain injuries can have delayed cognitive symptoms. AI users routinely settle in the first 30–60 days, sign a full general release, and then discover they have a serious injury that they can no longer recover for.

2. Undervaluing Injuries

Pain and suffering, loss of enjoyment of life, and emotional distress are real, compensable damages in California — often the largest component of a serious case. AI tools systematically understate these because they cannot evaluate the human reality of the injury or compare it to recent jury verdicts in your specific venue.

3. Missing Future Damages

A 35-year-old with a permanent partial disability has a lifetime of lost earning capacity ahead of them. That number, properly calculated by an economist with appropriate work-life and discount-rate assumptions, can dwarf the past medical bills. AI does not include this analysis, and adjusters will never volunteer it.

4. Failing to Identify All Liable Parties

Was the driver on the clock? The employer is liable. Was the vehicle defective? Product liability. Was the road negligently maintained? A government claim — with a 6-month notice deadline — may apply. Was there a dram shop or social host situation? Additional insurance may apply. AI almost never spots these alternative theories.

5. Accepting the First Offer

The first offer is almost never the best offer. Adjusters are trained to test claimants. AI users, lacking experience, often see a five-figure number and accept what should have been a six-figure case.

6. Damaging Recorded Statements

Within days of the crash, the at-fault driver’s insurance company will call and ask for a recorded statement. They will ask leading questions designed to get you to admit comparative fault, minimize your injuries, or commit to a version of facts you’ll later regret. You are under no legal obligation to give a recorded statement to the other side’s insurer. AI does not warn people about this. By the time clients reach me, the damaging statement is already in the file.

7. Not Understanding Liens

If your health insurance, Medi-Cal, Medicare, or a hospital paid for treatment, they have a legal right of reimbursement from your settlement. These liens can be aggressively asserted and substantially reduced — but only by someone who knows the rules. AI users frequently settle without addressing liens, then face collection actions for tens of thousands of dollars after the settlement money is gone.

Represented vs. Unrepresented Claimants: What the Data Actually Shows

The single most cited study on this question comes from the Insurance Research Council, which has repeatedly found that injured claimants represented by an attorney recover, on average, multiples of what unrepresented claimants recover for similar injuries. Even after deducting a contingency fee, the net to the claimant is typically far higher.

This is not because lawyers have a magical fee table the insurance company secretly honors. It’s because represented claimants:

• Develop the underlying case fully before demanding
• Document damages the way adjusters and juries actually evaluate them
• Identify all available coverage layers
• Present a credible litigation threat
• Negotiate with experience and patience
• Resolve liens favorably at the end

How Contingency Fees Actually Work

In a California personal injury case, you do not pay your attorney by the hour. You pay a percentage of the recovery — typically 33⅓% pre-litigation and 40% if a lawsuit is filed — and only if the attorney recovers money for you. If there is no recovery, there is no fee. The attorney also typically advances all case costs (filing fees, expert fees, deposition costs, medical record retrieval) and is repaid only out of the settlement.

Here is the practical reality. If a self-represented claimant settles a case for $20,000 that should have been worth $100,000, they keep $20,000. If a represented claimant settles the same case for $100,000, after a one-third fee and reasonable costs, they typically net $55,000–$65,000. The represented claimant’s net recovery is two to three times higher — even after the fee.

California-Specific Legal Complexities AI Cannot Navigate

California has some of the most plaintiff-friendly — and plaintiff-treacherous — personal injury laws in the country. A handful of these traps catch AI users almost every time.

Comparative Fault (Pure Comparative Negligence)

California is a pure comparative negligence state. You can recover even if you are 99% at fault — but your recovery is reduced by your percentage of fault. Adjusters routinely try to assign 20–40% comparative fault to claimants who did nothing wrong. Without an experienced advocate to push back, that allocation sticks.

Medical Liens and the Howell Rule

Under Howell v. Hamilton Meats, California limits a plaintiff’s recovery for past medical expenses to the amount actually paid (or the reasonable value of services, in some circumstances) — not the amount billed. The interplay between billed amounts, paid amounts, lien amounts, and recoverable damages is intricate. Get it wrong and you either understate damages or face a malpractice-level lien problem at the end.

Proposition 213

If you were operating an uninsured vehicle at the time of the crash, California Civil Code § 3333.4 — passed by voters as Proposition 213 — bars you from recovering non-economic damages (pain and suffering, emotional distress) regardless of fault. There are exceptions (DUI drivers, parked vehicles, employer-owned vehicles, and others), but they are technical. AI tools regularly fail to flag this issue, and self-represented claimants accept settlements that ignore the Prop 213 effect or, conversely, give it up when they shouldn’t.

Statute of Limitations

California gives you two years from the date of injury to file a personal injury lawsuit (CCP § 335.1). If a public entity is involved — a city, county, transit agency, school district, or the State of California — you must file an administrative claim within six months of the incident under Government Code § 911.2. Miss either deadline and your case is over, no matter how badly you were injured. AI tools do not calendar these dates for you.

Insurance Bad Faith

California recognizes the tort of insurance bad faith. When a carrier unreasonably delays, denies, or lowballs a claim — particularly your own first-party UM/UIM claim — they can be liable for compensatory damages, attorneys’ fees under Brandt, and in egregious cases punitive damages. Building a bad faith case requires very specific documentation of the carrier’s conduct over time. AI cannot construct this record.

Where AI Actually Can Help You (Used Correctly)

To be clear: I am not anti-technology. AI tools have a real, valuable role in the life of an injured person. They just are not a replacement for legal representation. Here are appropriate uses:

• Organizing your medical records, bills, and out-of-pocket expenses chronologically
• General education — understanding what comparative fault means, what a deposition is, what UM/UIM coverage covers
• Drafting questions to ask your attorney during the initial consultation
• Translating dense medical terminology in your own records into plain English
• Keeping a daily pain and symptom journal with consistent formatting
• Researching prospective attorneys — bar status, disciplinary record, trial experience, peer recognition

Used this way, AI can make you a better-informed and better-organized client. It cannot, and should not, replace an experienced California personal injury attorney.

If You Have Been Injured: What to Do Right Now

If you are reading this article because you have been hurt in a California car accident, here is my honest, practical advice — the same advice I give to friends and family.

First, get medical care and follow your doctor’s recommendations. Your health is more important than your case, and consistent treatment is also the foundation of a strong claim.

Second, do not give a recorded statement to the other driver’s insurance company. You are not required to. Politely decline.

Third, before you send any AI-generated demand letter, before you accept any offer, and before you sign any release, talk to an experienced personal injury attorney. The consultation is free. There is no obligation. And it is the single most valuable hour you will spend on your case.

For more information on related topics, see our pages on California car accident claims, how personal injury settlements are valued, uninsured and underinsured motorist claims in California, and what to do after a car accident. You can also learn more about Steven M. Sweat and our firm.

Frequently Asked Questions

Can I settle my car accident claim without a lawyer in California?

Legally, yes. There is no law that requires a represented claimant. Practically, it is almost never advisable for any case involving more than minor property damage. Insurance Research Council data — and three decades of my own experience — show that unrepresented claimants recover dramatically less, even after factoring in attorneys’ fees. The more serious the injury, the more disastrous self-representation tends to be. If you have any treated injury, talk to a lawyer before settling. Free consultations exist for exactly this reason.

Will insurance companies take me seriously without an attorney?

They will be polite. They will not take you seriously. Insurance carriers maintain detailed internal data on outcomes by representation status. They know an unrepresented claimant cannot file a lawsuit, take a deposition, or retain experts. Opening offers to unrepresented claimants are calibrated accordingly — often a fraction of true case value. The moment a credible attorney appears in the file, the reserve on your claim is typically increased and the negotiating posture changes.

Can ChatGPT write a valid demand letter?

ChatGPT can produce a document that looks like a demand letter. Whether it is valid — meaning effective at obtaining fair compensation — is a different question. A demand letter is only as strong as the investigation, evidence, medical documentation, damages analysis, coverage review, and credible litigation threat behind it. AI can do none of those things. Adjusters routinely identify AI-drafted demands and respond with lowball offers because they know there is no real case behind the letter. The document itself is not the problem. The absence of everything that should support it is.

How much does it cost to hire a California personal injury attorney?

Reputable California personal injury attorneys, including our firm, work on a contingency fee. You pay no money up front. You pay no hourly fee. The attorney advances the costs of building the case. If there is no recovery, you owe nothing. If there is a recovery, the attorney is paid a percentage of the settlement — typically one-third before a lawsuit is filed and a higher percentage if litigation becomes necessary. Even after the fee, represented claimants typically net more than unrepresented claimants for the same injury.

What is the deadline to file a personal injury claim in California?

The general statute of limitations for personal injury in California is two years from the date of the injury under Code of Civil Procedure § 335.1. If your claim is against a public entity (a city, county, transit agency, the State of California, etc.), you must serve a written government tort claim within six months of the incident under Government Code § 911.2. There are limited exceptions. Do not assume you have time. Talk to an attorney as soon as possible.

What if I already used ChatGPT to send a demand letter — is my case ruined?

Not necessarily. If you have not yet signed a release or accepted an offer, the case can usually be salvaged, though there may be ground to make up. The most important thing is to stop the bleeding now. Do not respond to the adjuster, do not give any further statements, and call an experienced personal injury attorney today for a free evaluation. The earlier we get involved, the more we can do.

About the Author

Steven M. Sweat is the founding attorney of Steven M. Sweat, Personal Injury Lawyers, APC, a Los Angeles-based California personal injury law firm. For thirty years he has exclusively represented injured individuals and the families of wrongful death victims throughout Southern California. He has been continuously selected to Super Lawyers since 2012, holds an Avvo 10.0 “Superb” rating, and is a member of the National Trial Lawyers Top 100 and the Multi-Million Dollar Advocates Forum. The firm handles every case on a contingency fee basis.

Steven M. Sweat, Personal Injury Lawyers, APC

11500 W. Olympic Blvd., Suite 400, Los Angeles, CA 90064

(866) 966-5240  |  ssweat@victimslawyer.com  |  victimslawyer.com

Disclaimer: This article is for general informational purposes only and does not constitute legal advice. Reading this article does not create an attorney-client relationship. Every case is different and outcomes depend on the specific facts. For advice on your particular situation, consult a licensed California attorney.