VIN accident check: what the records show, and what they never will

Vehicle history reports do not contain accidents. They contain records of accidents — which is a different thing, and the gap between them is where used-car buyers lose money. If a crash produced an insurance claim, a police report, a repair at a shop that sells its data to a commercial network, or a title event, it can appear. If it produced none of those, no provider on earth can show it to you.

That is not a reason to skip the checks. It is a reason to run the free ones, buy the paid one when the car justifies it, and then treat the result as a starting point for an inspection rather than a verdict that ends one.

• NICB VINCheck — free, and flags vehicles reported as total losses by participating insurers
• A free VIN decode — confirms the car is the trim and build the listing claims, which catches a different fraud
• NHTSA recall lookup — free, and unrepaired safety campaigns matter regardless of crash history
• State title tools where they exist, such as Florida's free FLHSMV check
• The seller's own paperwork — service invoices and insurance documents they may hand over without thinking

Commercial reports are worth buying when a car is expensive enough that a missed collision changes the deal materially. Their accident coverage comes from networks the providers built themselves, which is why coverage differs between them. Edmunds tested that directly by running both major products against one 2014 Porsche Boxster: the two came back with different service histories and even disagreed on how many owners the car had had, and neither was a superset of the other. Treat a single clean report as the absence of records rather than proof of an uneventful car. Our AutoCheck review and pricing guide cover which to reach for.

Separately and more cheaply, the federal NMVTIS record catches the most consequential accident outcome of all: the insurer write-off that became a title brand. At a few dollars, that check is worth running on any car you are serious about.

“Was it in an accident, and what did the repair cost?” is the question almost everyone asks, and it quietly assumes that one number decides the outcome. Across states, at least three different numbers can — and a seller can answer one of them completely honestly while the other two would have told you something else entirely.

The three are: what a shop estimated, what was actually spent, and what an insurer paid. They are not interchangeable, and states do not agree on which one governs.

Wisconsin’s salvage test runs on the estimated or actual repair cost, whichever is greater — so a seller quoting whichever of the two came in lower has not answered the state’s question at all. North Carolina reaches its brand two independent ways: repair cost and making the vehicle safe for the road exceeding 75% of fair retail market value, or an insurance company having paid a claim exceeding 75% of that value. The second route needs no repair estimate to exist. New Jersey, for a vehicle outside its eight-model-year branch and above its value floor, treats the insurer settling a total loss claim as a trigger in its own right. Alabama ties its test to a settlement actually paid on damage at or above 75% of fair retail value.

It is also worth knowing that a payment does not have to be a write-off to matter. Wisconsin brands a vehicle under seven years old that was transferred to an insurer on a claim for damages of 30% through 70% of its value — damage far short of salvage, and a permanent mark on the title. Our Wisconsin title check covers that band in full.

So the questions worth asking a seller are narrower and more specific than the usual one:

• Was an insurance claim ever paid on this vehicle — and roughly how much?
• What did the repair estimate come to, and was there more than one estimate?
• What was actually spent, and is there an invoice or a shop that can confirm it?
• Did an insurer ever take ownership of the car, even briefly?
• Was any claim paid to you, or to an owner before you?

None of those answers settles the brand question by itself, and they are not meant to. They tell you which record to go looking for, and whether the seller’s account is consistent with what you find. A seller who genuinely cannot say whether a claim was ever paid has told you something useful: that the conversation is not the check, and the national record is.

There is a column in the free federal data that almost nobody uses on this question. When an owner files a complaint with NHTSA, the record captures whether the failure involved a crash. That turns a pile of grievances into something an accident-minded buyer can actually read: not how many people complained, but what kind of thing went wrong.

Across the 777 model years in our data with enough complaints to measure, 4.3% of complaints involved a crash. That is the baseline. The useful part is that model years vary widely around it — 38 of them run at three times that rate or higher. A year sitting well above the line is not simply a noisy year. Its failures are the kind that end in a collision rather than an inconvenience.

That distinction is worth more than a raw complaint count. Two model years can carry identical complaint volumes while one is dominated by infotainment faults and the other by steering, braking or unintended acceleration. Only the second changes how carefully you want the car inspected, and only the second makes an unrecorded collision in its past more plausible.

Used that way it changes the order of work. Pull the year’s record free, see whether its failures skew toward crash-involved systems, and let that set how much you spend on the paid checks and how long you keep the car on a lift. Our model-year problem pages break the same data down by nameplate and year.

The single most useful thing to understand about accident records is that they are not produced by one system with uniform coverage. They come from two, and the two have opposite guarantees.

When a crash is bad enough to total the car, reporting is compulsory. Federal rules oblige insurers and the salvage trade to report, and state titling agencies feed the same federal system, so a total loss struggles to stay invisible. When a crash is bad enough to need a body shop but not bad enough to total the car, reporting is voluntary. A commercial network learns about the repair only if that shop happens to be one that reports to it.

Read the consequence carefully, because it inverts the usual advice. A clean federal title record is strong evidence that this car was never written off. A clean commercial report is weak evidence that it was never in a moderate collision, because the absence may only mean nobody chose to file. The two clean results are not equally reassuring, and treating them as interchangeable is how people talk themselves past a repaired quarter panel.

Picture the most common unreported crash: a driver with liability-only coverage backs into a pillar, gets an estimate, decides a claim would cost more in premium than the repair, and pays a body shop in cash. No insurer, no police, no title event, no reporting shop. The car is repaired, sold two years later, and every database on this page returns a clean result — accurately, because nothing was ever filed.

Multiply that by every driver with a high deductible and every fender-bender settled between two people at the roadside, and you have the reason a clean report and a good inspection are complementary rather than redundant.

• Panel gaps — run a finger along each seam; consistent factory gaps become uneven after repair
• Paint depth — a gauge costs little and reads refinished panels as noticeably thicker than factory
• Low raking light — stand at the corner of the car and sight down the flank for texture and color shifts
• Overspray — check rubber seals, wheel arches, the fuel filler and under the hood edges
• Bolt heads — a paint-cracked or rounded fender bolt means that panel came off at some point
• Date codes — glass, headlights and the radiator support carry dates that should roughly match the build

Where a report already shows a damage reported entry, it tells you which corner of the car to start on — which is most of what that entry is actually good for, since the phrase itself carries no legal definition and no measurement.

Modern cars remember things their dashboards do not display. A scan across every module — not just the engine — can surface stored and historical fault codes, airbag-system status, and modules whose part numbers or programming dates sit oddly against the car’s age. A replaced airbag control unit on a car with no reported accident is a conversation worth having before money moves. Any decent independent shop does this as part of a pre-purchase inspection; it is the cheapest sophisticated check available.

1. 1Decode the VIN — freeConfirm the car matches the listing, and see the factory build and open recalls.
2. 2Run NICB VINCheck — freeScreens theft and insurer-reported total losses in about a minute.
3. 3Buy the NMVTIS record — a few dollarsCatches title brands and write-offs, including ones washed across state lines.
4. 4Add a commercial report if the car is expensiveTheir accident networks hold records the federal system never sees.
5. 5Inspect, gauge and scanThe step that finds the unrecorded repairs — and the only one that describes the car today.