I build AI extraction for insurance documents, and a while ago I wrote up what I’d learned about the property side — the coverage letters as an onion, the form numbers as a ladder.
Then I moved to auto, and got stuck immediately on two words: at-fault and no-fault.
I read the definitions three times. They kept sliding off. And while I was failing to absorb them, I discovered something more embarrassing: I had the basic sequence of events wrong too. If you’d asked me what happens after a crash in the U.S., I’d have said — police come, police determine who’s at fault, then you call your insurer.
That is not what happens. A police report does not decide liability. The officer documents the scene, writes a narrative and a diagram, and may issue a citation. A citation is powerful evidence of negligence. It is not binding.[^1] The people who actually decide fault are the claims adjusters at the two insurance companies, negotiating with each other, often landing on a percentage — 70/30, not guilty/innocent. If they can’t agree, a jury decides, and only then is anything final.
I’d imported the wrong model wholesale. And once I noticed that, the more interesting question became: why is the process shaped like that? Why is it a negotiation between two companies rather than a determination by an authority?
The answer turned out to be a single question that runs underneath all of it, and it isn’t about cars at all.
Count the signatures
An insurance policy is a contract. Contracts have signatories. Count them:
| Who | |
|---|---|
| First party | the insured — you |
| Second party | the insurer |
| Third party | anyone who didn’t sign this contract |
That’s the whole derivation. The numbering isn’t jargon, it’s just counting.
Now notice something odd: you will basically never hear anyone say “second party.” Not in a claim, not in a policy, not from an adjuster. First party, constantly. Third party, constantly. The second party is structurally invisible — because in any conversation about a claim, the insurer is the room you’re standing in. It doesn’t need a number. It’s the one you’re talking to.
So a perfectly sensible three-item list has worn down to a strange-looking pair, and generations of people have learned “first-party claim” and “third-party claim” as opaque idioms without ever being told they’re ordinals with a gap in the middle.
Here’s the whole thing in one line: a third party is someone who didn’t sign it.
Which gives you two kinds of claim:
- First-party claim — you claim on your own policy.
- Third-party claim — someone claims against your policy (they’re eating your liability coverage), or you claim against theirs.
That’s it. That’s the seam. Everything below is consequences.
A free reading rule, before we go on
The seam pays for itself immediately, because one word gives it away:
When
liabilityappears in the name of a coverage, it’s the third-party side.
Bodily Injury Liability, Property Damage Liability, Personal Liability — all third-party, no exceptions worth worrying about. It’s the cheapest piece of insurance literacy I’ve picked up.
One thing to pin down, though, because it flips: the liability coverage on your policy covers your responsibility to other people. So if someone rear-ends you and you claim against their liability coverage, you are the third party in that transaction. The label doesn’t mark a person — third party is always just whoever didn’t sign the contract in question, and which contract you’re holding decides who that is.
And here’s the counterexample that convinced me the rule was real rather than a coincidence of naming. There’s a pair of auto coverages for when the driver who hit you had no insurance, or nowhere near enough: uninsured motorist and underinsured motorist. They’re entirely about the other driver’s responsibility — but they live on your policy, and they’re first-party.
Look at the names: Uninsured Motorist, Underinsured Motorist. No “liability” in either one. For an industry that names things as badly as this one does, that’s a surprisingly honest piece of drafting — the word stayed away, because the responsibility being discussed isn’t yours.
The seam decides who is owed something
Here’s why this isn’t a vocabulary lesson.
When you claim on your own policy, the insurer owes you a duty of good faith and fair dealing. It’s your contract. If they stall, lowball, or fail to investigate properly, that can be bad faith — and you can sue them for it.
When you claim against someone else’s policy, that insurer owes you essentially nothing. Its duties run to its own insured. You’re a stranger who showed up asking for money out of their customer’s coverage.[^2]
Sit with that for a second, because it rearranges how the whole process reads:
The other driver’s adjuster isn’t being difficult because adjusters are unpleasant people. They owe you nothing, and their actual job is protecting their customer from paying too much. Pushing back on you is the assignment.
Behaviour that might be illegal on a first-party claim is, on a third-party claim, just Tuesday.
So a third-party claim is adversarial by construction. Not by temperament, not by company culture — by the structure of who signed what. And that answers the question I got stuck on earlier: the process after a crash is a negotiation between two companies because nobody in it owes the injured person a duty. Each insurer is representing its own signatory. There’s no authority in the room, only two representatives.
(One precision, because I got this wrong myself at first: there is a thing called third-party bad faith. But it belongs to the insured, not to you. It’s for when your insurer refuses a reasonable settlement inside your limit and leaves you personally exposed to a bigger judgment. The claimant standing outside the contract still generally has no such claim.[^2])
A liability policy sells two things, and people only count one
While we’re on the third-party side — a detail that took me a while to notice, and which changes what a liability limit even means.
When you buy liability coverage, the insurer promises two separate things:
- Indemnify — pay the damages, up to your limit.
- Defend — hire and pay lawyers to fight the claim on your behalf.
And in standard personal-lines and commercial general liability policies, defense is paid in addition to the limit — it doesn’t eat the money available to pay the claimant.[^3]
Think about what that’s worth on a small policy. If you carry a minimum limit and someone sues you for ten times it, the defense promise can easily be the more valuable half of what you bought. Roughly speaking, a good chunk of a liability premium buys “somebody will show up in court for you” — and that half doesn’t appear anywhere on the declarations page as a number.
Worth knowing that it isn’t universal: some policies (professional liability especially) pay defense inside the limit, so every dollar of lawyer is a dollar less for the claimant. Two policies can show the identical limit and be materially different products.
This also tells you where the money is in the next section. When a jurisdiction takes away the right to sue, it isn’t only saving claim payments — it deletes the entire defense cost line.
So: no-fault
Now the two words that stopped me.
No-fault does not mean nobody was at fault. This is the single most misleading name in the business — and I say that having already written a whole article about how bad insurance names are. Fault does not go anywhere. Police still cite. Adjusters still assign responsibility. It still lands on the at-fault driver’s record. Their premium still goes up.
What “no” negates is much narrower: when paying your medical bills, fault is not consulted.
And with the seam in hand, you can say what’s actually happening, which the name never will:
No-fault moves the injury claim out of a contract you never signed and into one you did.
That’s the entire mechanism. In an at-fault state, your injury claim goes against the other driver’s policy — third-party, no duty owed to you, adversarial, slow, and gated on establishing liability first. In a no-fault state, your injury claim goes against your own policy, to a company that owes you good faith and has your name on the paperwork.
The speed isn’t magic and it isn’t generosity. It’s a change of counterparty. “Fault isn’t consulted” is the consequence of the relocation, not the cause of the speed.
Two things fall out immediately once you see it this way:
It only covers people, not cars. No-fault handles bodily injury. Vehicle damage still runs on at-fault rules — you still need the other driver’s property damage liability, or your own collision coverage. So one accident runs two payment logics in parallel, on opposite sides of the seam. Florida takes this to its logical end: it mandates property damage liability but no bodily injury liability at all, because the people-half is delegated to your own policy and the car-half isn’t.
It’s a trade, and the price is your right to sue. You get paid fast without proving anything; in exchange, you can’t sue for pain and suffering unless your injury clears a statutory threshold. Fast money for surrendered litigation. That’s the deal, and it’s why “no-lawsuit” would have been the honest name.
A non-U.S. detour, clearly markedEverything above is about the United States. But I live in British Columbia, and BC happens to be the most extreme version of this trade anywhere nearby, which makes it a useful thing to look at — so, flagged as an aside and not part of the U.S. picture.
In 2021 BC moved to a system called Enhanced Care. Care and recovery benefits went to $7.5 million, up from $300,000, available regardless of fault — and the right to sue an at-fault driver was removed by statute, with a narrow exception for drivers convicted of certain criminal offences. The province reported saving roughly $1.5 billion in legal costs, which is what funded the benefit increase and a premium reduction.[^4]
Two things worth pulling out.
First, the naming. “Enhanced Care” is one of the rare cases where the industry noticed it had a bad name and fixed it — it describes what you get instead of describing the abstraction it negates. Compare “no-fault,” which tells you nothing and actively misleads.
Second — and this is the whole point of this article, said by the insurer itself. From the public overview of the program: “fault still matters and accountability remains a key tenet of this insurance model. [The insurer] still determines responsibility after a crash, which can impact driver records and insurance premiums.”[^4]
There it is, from the organization that runs a no-fault system: fault is still determined, and it still hits your premium. If you ever wondered how you can be found zero percent responsible and still see your renewal go up — that’s a different question with several possible answers, but “no-fault means fault stopped existing” isn’t one of them.
The critique of BC’s version is also instructive, and it lands exactly on the seam: moving everything inside your own contract means the amount is now set by an adjuster, and you can no longer take that number to a court. Lawyers there have called the $7.5 million figure a red herring for that reason.[^4] Relocating a claim across the seam buys speed, and it costs you the outside referee. Both directions of that trade are real.
The same seam cuts the house
Here’s the part I didn’t expect, and the reason I think this is the single most useful concept I’ve picked up doing this work.
It’s the same seam in property insurance. Different vocabulary, one division:
| First-party side (your stuff, your body) | Third-party side (other people) | |
|---|---|---|
| Homeowners | Section I — Coverage A dwelling, B other structures, C personal property, D loss of use | Section II — Coverage E personal liability, F medical payments to others |
| Auto | collision, comprehensive, PIP, medical payments, uninsured/underinsured motorist | bodily injury liability, property damage liability |
A homeowners policy is split into Section I and Section II, and that split is the first-party/third-party seam. In the property article I described that division as the seam the whole market is cut along, without being able to say what it was made of. This is what it’s made of.
And three facts I’d previously just memorized as brute structure turn out to be derivable:
Section II is identical across every homeowners form. HO-1 through HO-8 differ enormously in how they treat your building — named perils, open perils, replacement cost, actual cash value. The liability section doesn’t move at all. Of course it doesn’t: your promise to strangers doesn’t depend on how broadly you insured your own house. Different side of the seam, unaffected.
The dwelling fire forms have Section I and no Section II. A DP policy is a pure first-party contract — it protects the building and nothing else. That’s why a landlord using one has to buy liability back separately. Not an oversight; the form only lives on one side of the seam.
And the one that finally convinced me the seam is real: Coverage F. Coverage F is medical payments to others — it pays medical bills for someone who gets hurt at your house, with no lawsuit and no finding of fault. It pays medical bills, exactly like auto medical payments coverage does.
Auto medical payments is first-party. Coverage F is third-party.
Same kind of loss, opposite sides. Because the test was never what kind of damage is this — the test is whose contract is being claimed against. Coverage F pays a non-signatory. Auto medical payments pays you. That single difference puts them in different halves of the policy.
Which gives you a rule that works everywhere, and it’s the thing I’d keep if I had to throw out the rest of this article:
To place any coverage, don’t ask what it pays for. Ask whose contract is being claimed against.
Where I’ve landed
(Follow-up, since published: I walked every coverage on an auto policy through this seam, one at a time, then ran a single $180,000 injury through all nine of them — which showed something I hadn’t expected about which coverages are compelled and which ones actually pay you. That’s here.)
The onion and the ladder were about how much is covered. This is about whose contract does the covering — and it’s the more load-bearing of the two, because it’s the same in every line, while form numbers are a separate vocabulary per line.
It’s also the answer to the thing that started this. The reason a U.S. crash resolves as a negotiation between two companies rather than a ruling by an authority is that the injured person is standing outside the contract that’s supposed to pay them. Everything downstream — the adversarial adjuster, the slowness, the percentages, the entire invention of no-fault to get around it — is that one structural fact working itself out.
Three ordinals. One of them never gets said out loud. And it reorganizes the whole business.
Notes and sources
- [^1] On police reports and fault: multiple U.S. personal-injury practices state plainly that a police report doesn’t have legal authority to determine fault — liability is decided by insurance adjusters or, in litigation, by the trier of fact. A citation is treated as strong evidence of negligence but is not binding in a civil claim.
- [^2] On the duty asymmetry: an insurer’s duties in a third-party claim run to its own insured, not to the claimant, so the claimant generally cannot bring a bad-faith action. First-party bad faith (your own insurer mishandling your claim) is the established category. “Third-party bad faith” is a real doctrine but belongs to the insured — typically for an insurer’s failure to settle within limits, exposing its own policyholder to an excess judgment.
- [^3] On defense costs: in the standard commercial general liability form, defense expenses are covered under Supplementary Payments and do not reduce the policy limits; personal-lines liability works the same way. Some policies — professional liability in particular — are written with defense inside the limit, which is a materially different product at the same stated limit. Courts also commonly observe that the duty to defend is broader than the duty to indemnify, though how much broader depends on the specific policy language.
- [^4] On British Columbia: Enhanced Care launched 1 May 2021; care and recovery benefits rose to $7.5M from $300,000; the right to sue an at-fault driver was removed by statute except where the at-fault driver is convicted of certain Criminal Code offences; the provincial insurer reports roughly $1.5B in legal-cost savings funding the change. The quotation on fault still mattering is from the insurer’s own public overview of the program. The “red herring” critique of the $7.5M figure — on the grounds that the amount is set by an adjuster and can no longer be tested in court — was made by B.C. personal-injury counsel in local business press coverage of the transition.
I’m learning this in public and I’d rather be corrected than confident. The property↔auto unification in the last section is my own reading rather than something I got from an expert — if you work in this and it’s wrong, please tell me.