Inspecting a used car: what to check, in what order, and what ends the viewing

Most guides to inspecting a used car are written in the order that sets the price. This one is written in the order that ends the viewing.

That distinction is the whole argument, and the piece this is translated from gets it wrong in a way worth naming, because the mistake belongs to the market rather than to one author. It states outright that a repainted body can never be put right while the mechanical side always can, and then spends its first three thousand words on paint. As a statement about resale value that is more or less true, and it is the seller's ordering. As a buyer's ordering it is backwards. A resprayed wing is cosmetic. The question that matters is whether the structure under it was straightened properly, and paint cannot answer that in either direction: a car with perfect panel gaps and a badly pulled front end looks better than an honest car with a scuffed bumper.

So what follows starts with the checks that cost nothing and can end a viewing in ten minutes, and works down to the ones that only matter once you already want the car. The test drive is about two-thirds of the way down. Most people do it first.

The original is the most-read piece in this archive by a wide margin, and it is genuinely useful — it is thorough about the body and the interior in a way most guides are not. It also contains one instruction that destroys the evidence you came to find, one engine test whose reasoning is wrong in three separate places, and a clutch test that on a front-wheel-drive car does not test anything. All three are dealt with where they arise.

Before you look at a single car

The cheapest decision in the whole process is the one made before any car is involved: which model, on grounds that have nothing to do with the individual car's condition.

The archive makes this point once, in a different article, and then never returns to it — and it is the best paragraph in either piece. Its example was local, but the reasoning is universal. It argues that one particular saloon is the lower-risk used buy because it shares most of its parts with two other high-volume cars, because production continued for years afterwards, because every parts shop in the country stocks for it, and because every mechanic has worked on one. None of that is about the car in front of you. All of it determines what the car costs to own.

Two things follow that are worth doing from a chair:

Find out whether the engine is an interference design. In an interference engine the valves and the piston pass through the same space at different moments, and the timing belt is the only thing keeping that appointment. If the belt breaks, they meet. Most engines on the road are built this way — the usual figure quoted is around seventy per cent — and the consequence is that a missing belt-change record means two entirely different things depending on the answer. On a non-interference engine it is the price of a belt. On an interference engine it is the price of a cylinder head, and you should assume you are buying one. The original garbles this badly, opening a sentence about the alternator belt and closing it with the valves hitting the pistons; those are two different belts and only one of them can do that.

Find out what this model is known for. Every model has two or three failures that are effectively scheduled. Knowing them turns a vague inspection into three specific questions.

And a caution against over-reading the model's reputation, which the archive puts better than most: within a single model, individual cars vary enormously, and everybody knows somebody whose example was faultless and somebody whose example was a disaster. The model tells you what to check. It does not tell you what you are looking at.

Get there before the engine is warm

This is the largest single omission in the original, which never mentions it, and it is the one thing on this page that has to be arranged in advance rather than noticed on the day.

A cold engine tells you more in the first thirty seconds than the rest of the viewing put together. Almost everything that is expensive announces itself on a cold start and then goes quiet: the top-end rattle before the oil pressure comes up, the blue puff from worn valve stem seals, the diesel that needs a long crank, the hydraulic tappet that clatters for ten seconds, the exhaust blow that seals as the manifold expands, the coolant leak that stops weeping once everything has swelled. Run the engine for fifteen minutes and every one of them disappears.

So: ask for the car to be cold when you arrive, and say why. A seller with nothing to hide will not mind.

If it is warm when you get there, that is itself a finding. It does not prove anything — people move cars, people go to work — but you have lost the most informative thirty seconds of the day, and you should either arrange to come back or price the car as though you had not seen it start. A seller who has "just been out to fill it up" for the second viewing as well has told you something.

Take the temperature question seriously in the other direction too, later on: the car needs to reach full operating temperature during the drive, because that is when the cooling system and the radiator fan get tested at all.

The dashboard, in the first ten seconds

Turn the ignition to the position before the starter — everything live, engine not running — and watch the whole cluster light up.

Every warning lamp is supposed to illuminate at that moment. It is a deliberate bulb check, and it exists so that a lamp cannot fail silently and leave you driving a car whose warnings have no way of reaching you. Start the engine and they should go out.

The important part is the failure that reads as good news:

A warning lamp that never lights at all is worse than one that stays on. A lamp that stays on is a fault somebody has not fixed. A lamp that never appears is a fault somebody has hidden — the bulb pulled, the LED unsoldered, or the cluster swapped for one from a healthier car.

Learn the full set for the model beforehand if you can, because you cannot notice the absence of a lamp you did not know to expect.

The airbag lamp above all. It should light for roughly three to six seconds and then go out. That interval is the restraint module running its self-test and confirming it can reach the lamp; if it finds a fault afterwards it has a working way to say so. A car whose airbag lamp never appears may have no functioning restraint system at all — and this is the single most likely thing to have been tampered with on a car that has been crashed, which is precisely the car the rest of this inspection is trying to identify. The original spends a long passage on airbags and confines itself to checking that the cover sits flush and has no bulges. That checks whether the airbag is there. It does not check whether it works.

The same reasoning applies to the ABS and engine-management lamps, for the same reason and by the same method.

Plug something in

A code reader is the cheapest tool in this article and the one the original was written too early to use. It is worth more than every paint gauge on the market.

Read the stored codes, the pending codes and the freeze-frame data. A stored code names a fault. A pending code names one the car has seen once and is waiting to confirm, which is often the fault that has not yet become a symptom. Freeze frame records the conditions at the moment the code set. If you get a code number, the corpus on this site has a page for most of them — P0300 for a misfire, P0171 for a lean bank, P0420 for a tired catalytic converter, P0128 for a thermostat stuck open. Reading live data covers what to do with the rest.

Then look at the readiness monitors, which is the part almost nobody does. These are the self-tests the engine management runs on its own emissions systems as you drive — catalyst, oxygen sensors, evaporative system, EGR and so on — and each reports as complete or incomplete. Clearing the fault memory resets every one of them to incomplete, and so does disconnecting the battery. They only return to complete after the car has been driven through the specific conditions each test needs, which typically takes a number of trips.

So:

No stored codes, but several monitors incomplete, means the memory was cleared recently. Not necessarily this morning. But recently, by somebody, on a car being sold — and a clean scan on a car whose monitors have not finished proves nothing at all.

That combination is one of the strongest single tells available at a viewing, it takes fifteen seconds, and no amount of looking at the car can substitute for it. It is also worth knowing in your own favour: if the seller genuinely disconnected the battery last week for an unrelated reason, that is a normal explanation and easily checked against everything else you have seen.

A scan tool also reads mileage from more than one place. On most modern cars the odometer reading is written into several modules — the instrument cluster, the engine ECU, the ABS unit, sometimes the keys — precisely so that a rollback has to be done in every one of them to be consistent. A cluster reading well below what the other modules hold is not ambiguous.

That is the modern version of a check the original does at length and entirely by eye: wear on the pedal rubbers, the steering wheel, the driver's seat bolster and the gear knob, plus its suggestion to look for fingerprints on the inside of the instrument glass. The physical checks are still worth making — they are the ones that catch a car whose whole cluster was swapped — but the fingerprint test belongs to mechanical odometers, and the electronic ones are read, not inspected.

The engine bay: do not let anybody wash it

Here is the instruction in the original that costs the most:

It advises that an honest seller has the engine bay and the underside steam-cleaned before showing the car, and that if you find a dirty engine you should ask for it to be washed so you can see the components clearly.

Do not do this. Oil weep trails, coolant crust, the powdery residue around a leaking gasket and the wet film on the bellhousing are the evidence. Washing them off removes exactly what you came to read, and it removes it in a way that cannot be undone during the viewing.

An evenly grimy engine bay is a readable document. It is not a sign of neglect on a car of any age, and a uniform layer of road film everywhere is close to the best case: there is nothing standing out because nothing is leaking hard.

The pattern to look for is the opposite one. A bay that is spotless on top and wet underneath is the specific tell, because cleaning from above never reaches the bottom of the block, the sump seam or the bellhousing. So:

• Take a torch and look at the underside of the engine and gearbox, the faces pointing at the ground. That is where a leak collects regardless of what has been cleaned above it.
• Look at the ground where the car has been standing, and again after it has been running for ten minutes. Fresh drops show up on dry tarmac long before they build into a stain.
• Note that engine cleaning is not inherently dishonest — it is the standard first step in diagnosing a leak, precisely because a clean engine makes a fresh leak obvious. That is why it is worth asking when it was done and what for.

While the bonnet is up, the routine checks are worth doing in the order that costs least: the coolant in the expansion tank should be a clear colour rather than rusty or oily; a mayonnaise-like emulsion under the oil cap or in the coolant is the classic head gasket sign, though a car used only for short journeys can produce a little of it innocently; the oil on the dipstick should be oil rather than a gritty suspension; the engine mounts, hoses and auxiliary belt should not be perished or glazed.

The oil filler cap test, which the original gets wrong three times over

The original describes a real test and then misreads it comprehensively. Its version: with the engine running, take the oil filler cap off, put your palm over the opening, and if you feel air pressure or see blue smoke, walk away — because it shows "compression in the cylinder head", meaning the engine has been rebuilt.

The test is worth doing. Every part of that reasoning is wrong.

First, what a healthy engine does there is not nothing — it is suction. On an engine with a working crankcase ventilation system, the crankcase is held at a slight vacuum at idle, and taking the cap off is itself a small vacuum leak that may disturb the idle. Gentle pulsing is normal. Steady positive pressure, visible vapour pushing out, or a cap that will not sit still is what indicates a problem.

Second, pressure there has two causes and only one of them is expensive. It means combustion gases are getting past the piston rings faster than the ventilation system can clear them — worn rings or bores, which is an engine rebuild — or it means the PCV valve or breather is blocked, which is a cheap part and an afternoon. Those produce the same symptom. Walking away on the strength of the symptom alone means walking away from cars with a five-pound fault.

Third, it has nothing whatever to do with a skimmed cylinder head. Crankcase pressure is generated below the pistons. A head that has been machined flat is a different repair in a different part of the engine, and the original's inference connects two unrelated things.

And the test that actually answers the question is a compression test — or better, a leak-down test, which distinguishes rings from valves from a head gasket rather than just reporting a low number. Neither is a viewing-day test on somebody else's car, which is the honest position: the filler cap tells you whether to pay for a proper inspection, not whether to buy.

The battery, and the green dot that means less than it looks

The original says to check that a sealed battery's indicator is fully green, and separately that a healthy battery proves itself if the engine starts and stops a couple of times.

Neither is a test. The indicator is a hydrometer floating in a single cell — usually the middle one — and it reports state of charge, not capacity. It says nothing about the other five cells and nothing about whether the battery can still deliver cranking current on a cold morning, which is the only property anybody actually cares about. A battery that is 40% down on capacity but fully charged shows a healthy green dot.

And starting a warm engine twice is the easiest thing a dying battery does. The real test is a load or conductance test, or at minimum watching the voltage while it cranks — using a multimeter covers the method and the battery article covers what the numbers mean. On a viewing, treat the battery as an unknown and a small negotiating item rather than as something you have checked.

The body: what paint tells you, and what it does not

The original is at its best here, and this section keeps most of its advice: look at the car in daylight, outdoors, from a shallow angle so that reflections run along the panel; check that the gaps between the doors, bonnet and boot lid are even and consistent side to side; open and shut each door listening for a clean latch rather than a metallic clash; look down the sills and inside the door shuts; lift the boot floor; check the pillars and roof, and treat any repair there as a different conversation entirely.

The one thing to add to all of it: panel gaps and paint tell you where the car was hit. They do not tell you whether the repair was any good, and that is the question. A structural repair done on a jig by somebody competent leaves a car that is safe and worth buying. The same damage pulled roughly and filled leaves a car that will never track straight and will fold unpredictably next time. Both can be under paint that matches.

The paint gauge, and why the original's version misleads

The original describes a pen-shaped tool it calls a magnet test, and gives the reading as a yes or no: if it releases from the panel easily there is filler under the paint, and if it comes away with some resistance the paint is "probably" original.

The tool is real — it is a magnetic pull-off gauge, and it works by measuring the force needed to detach a magnet, which falls as the coating between the magnet and the steel gets thicker. Three corrections, and the last one matters most:

It only works on steel. Magnetic gauges read nothing meaningful on aluminium or plastic, and modern cars use both extensively — bonnets, wings, boot lids and tailgates are frequently aluminium, and bumpers are always plastic. On an aluminium bonnet the magnet behaves exactly as it does over deep filler, which means the original's decision rule condemns an undamaged panel. A gauge that handles both reads aluminium by eddy current instead, and the cheap ones do not.

The output is a number, not a verdict. Coating thickness is measured in microns. Factory paint on a modern car is typically somewhere around 100 to 150 microns in total, though it varies by manufacturer, model and even between panels on the same car. A repainted panel commonly reads 200 or more; filler under paint reads far higher again.

And the reading that matters is the comparison, not the absolute. Because the factory figure varies so much, a single number in isolation is nearly meaningless. Walking the gauge over every panel on one car and finding that one wing reads twice what the roof and the opposite wing read is conclusive; finding 180 microns on a car you have not measured elsewhere is not. This is why the tool is useful and the original's pass-or-fail reading of it is not.

The tyres are the cheapest instrument on the car

The original mentions tyre wear once, in a closing checklist, and never says what it means. That is a real loss, because the tyres are a written record of the suspension and alignment, and they cannot be cleaned up for a viewing.

• Wear concentrated on one shoulder across an axle points at alignment — see wheel alignment for which angle does which.
• Feathered edges you can feel by running a hand across the tread points at toe.
• Cupping or scalloped patches points at worn dampers letting the wheel bounce.
• Different brands across one axle, or a nearly-new tyre on one corner, is worth a question. Sometimes it is a puncture. Sometimes it is a kerb strike that also bent something.
• The date code. A four-digit stamp in an oval gives the week and year of manufacture. Rubber hardens with age regardless of tread depth, so a low-mileage car can be on tyres that are finished. Tyres covers the full sidewall.

Budget for a set if anything above is true. Then look at the brake discs through the wheels for a pronounced lip or heavy scoring — brake pads and discs covers what is normal wear and what is not.

The drive

Drive it yourself, for at least twenty minutes, on a route you choose rather than the one round the block the seller suggests. It should include a rough surface, a stretch at speed, and enough time for the engine to reach full temperature.

Radio off, windows down for the first few minutes. Almost every finding on a test drive is a noise, and the two most common ways to miss one are a stereo and a closed cabin.

On the pull test, do not do what the original says. Its instruction is to take both hands off the wheel at 100 km/h on a motorway. That is a test performed at the highest available speed, in traffic, in a car whose brakes, tyres and suspension you have not yet verified — and it is the reason the car is being tested in the first place. Do it at 40 to 60 km/h, on a flat straight with nothing around you, and relax your grip rather than removing your hands. Note also that road camber drains water toward the kerb, so almost every car drifts gently that way on a normal road: a slight pull is the road, not the car. A definite one is worth chasing.

Brakes. Brake firmly from speed on an empty road, once you have checked what is behind you. You are feeling for pull to one side, for a pulsing through the pedal that indicates warped or unevenly worn discs, and for a pedal that sinks slowly under steady pressure, which is a hydraulic problem rather than a wear problem. Note that a car that has been standing will have surface rust on the discs and will feel poor for the first few stops. The original's version of this test — pressing the pedal a few times while stationary — finds only the worst failures.

The clutch, which is the other test the original gets wrong. Its version: handbrake on, select fourth or fifth, and hold the clutch at the biting point; if the engine stalls the clutch is good.

Two problems.

The handbrake alone does not hold a front-wheel-drive car. It acts on the rear wheels, which on most cars are not the driven ones — that is why it can be used while parked without loading the transmission. Applying it and driving against it simply drags the rear tyres. The car in the original's own worked example is front-wheel drive, and it says so elsewhere on the same page.

And holding the biting point is how you destroy a healthy clutch. Slipping a clutch against a held driveline is the maximum-heat condition for the friction surface; it is the failure mode being tested for, applied deliberately.

The correct version is short and done once: handbrake on and firm pressure on the footbrake, third gear, raise the clutch briskly and fully. A healthy clutch stalls the engine within about a second. If the revs rise and the engine keeps running, the clutch is slipping. Do it a single time — this is a hard test on the component even when done properly, and the car is not yours yet. The clutch article covers what the pedal feel and biting point height tell you separately.

The gearbox. On a manual, take your hand off the lever after each change and see whether it stays in gear; feel for a synchro that grinds on a fast second-to-third; check reverse. On an automatic, you are feeling for a delay between selecting drive and the car taking up, for a shift that arrives as a thump, and for hunting between ratios at a steady throttle. A CVT has its own set. Any of them is expensive, and all of them are easier to feel from cold, which is the other reason for arriving before the engine is warm.

Suspension and steering. Over a rough surface, listen for knocks from below rather than watching the ride quality — a car can ride badly and be sound, or ride acceptably on worn joints. Vibration through the steering at a particular speed is usually wheel balance; vibration through the seat is more often driveline.

Temperature and smoke. Once the engine is fully hot, check the gauge sits where it should and that the radiator fan cuts in and out. Then, with somebody watching behind, hold the revs and let them drop: blue smoke is oil, black is over-fuelling, and thick white that persists once the engine is hot is coolant. Light white vapour on a cold morning is condensation and normal — the original is right about that, and right that it should stop once the engine is warm.

The paperwork, and the one thing that can cost you the car after you have paid for it

The original has a long, careful section on documents, and almost all of it is about establishing that the car is what it claims to be: matching the VIN and engine number on the car against every document, checking that the identity plates have not been disturbed, checking the registration document is genuine. All of that transfers, and it is the right instinct. Match the VIN in every place it appears — the plate, the screen, the door aperture, the paperwork — and treat any disagreement as the end of the viewing rather than as an administrative problem.

What it never mentions is the failure mode that costs the most:

Check whether there is finance outstanding on the car. Under a hire-purchase or conditional-sale agreement the lender owns the car until the agreement is settled, and the seller has no right to sell it. Where that claim survives the sale — and whether it does turns on local law and on whether you can show you bought in good faith — you can pay the seller in full, take the car home, and have it repossessed. You lose the car and the money, and your recourse is against a person who has already gone.

Most markets have a paid check that covers this along with whether the car has been recorded as stolen or previously written off, and it costs a small fraction of a percent of the purchase price. It is the single highest-value thing on this page per unit of effort.

The rest of the seller check is more prosaic than the original's version, which advises getting to know the owner personally on the theory that a person treats their car as they treat themselves. That is a nice line and it is not usable: the tidiest, most reassuring sellers include the traders posing as private individuals, which is a specific and common arrangement that exists to strip the buyer of the protections a trade sale carries. The checkable version is duller and works: does the name on the registration document match the person in front of you, and are you at the address on it? Selling a car from a layby, or from a house where three other cars are for sale, answers the question on its own.

What ends the viewing

The title of this piece promises an order, and the order is only useful if some findings are allowed to end it. These are judgements rather than rules, but the split is real: some findings are prices, and some are answers.

And the finding that is not on the list, because it applies throughout: a seller who resists any of this. Every check above is either free or nearly free, none of them harms the car when done properly, and a private seller with a sound car has no reason to object to a cold start, a code reader or twenty minutes on a route you picked. Refusal is information, and it is the cheapest information available all day.

Video guides

Video, for the parts of this that are easier watched than read.