I have spent over 30 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. Trucking companies have to keep electronic logging device (ELD) records for only six months, so they need to be requested early. 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.

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, including whether the claimant had a lawyer. 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.

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 in defense-attorney and expert-witness fees, and a case that goes to trial can cost the insurer a great deal 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

A 1999 Insurance Research Council survey found that injury claimants who hired a lawyer received total payments about 3.5 times higher, on average, than those who did not. That figure is before attorney’s fees, and represented claimants in the survey also tended to have more serious injuries, so it isn’t a promise about any one case. 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 a share of the 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. The same bar applies to an injured driver convicted of DUI for the crash, and there is a narrow exception: an uninsured owner injured by a driver who is convicted of DUI can still recover non-economic damages. The details 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, and uninsured and underinsured motorist claims in California. You can also learn more about Steven M. Sweat and our firm.

Frequently Asked Questions

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 over 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.