Nearly 3 in 4 Car Noise Complaints to NHTSA Report No Warning Light

In July 2026 the owner of a new Chevrolet Silverado 1500 filed a complaint with the federal government. At about 3,000 miles, they wrote, "the engine began making a loud knocking noise that became progressively worse while driving." The truck was diagnosed as needing a complete engine replacement. Then the line this study is about: "No warning lights, no check engine light, nothing." (ODI 11751681.)

That owner is not unusual. On 13 September 2026 we downloaded every vehicle complaint NHTSA's Office of Defects Investigation has published for complaints received since 1 January 2015, 777,040 distinct complaints, and read the owner's description in each one by published rules. 64,831 describe a sound. In 47,707 of them, 73.6 percent, the owner reports no warning light, no trouble code and no dashboard message. 6,184 go further and say in plain words that nothing came on, and one in six of those describes an engine that failed.

I built Pulscar — a tool that listens to a 30-second recording of an engine — after a mechanic charged me $380 to "diagnose" a knock that turned out to be a loose heat shield. That is my bias, and it is why every rule, count and table behind this page is published for anyone to re-run.

The short version

What did Pulscar's analysis of NHTSA complaints find? That most drivers who hear a problem do not report being warned about it. Across 777,040 vehicle complaints received from January 2015 to 8 September 2026, 64,831 describe a sound in the owner's words. In 73.6 percent of those, 47,707 complaints, the text mentions no warning light, trouble code or dashboard message, and 6,184 state outright that none appeared. That explicit group is not a pile of minor squeaks: 17.2 percent of it describes an engine that failed, seized or needed replacing, against 3.6 percent of all complaints, and 9.5 percent records a tow. The quiet share is above 80 percent for structure, suspension, wheels and steering. It is highest of all for cars two years old or newer, at 79.1 percent, and electric cars are no exception, at 72.6 percent.

How did Pulscar decide whether a complaint reported a warning light? By reading the owner's own text with published rules, in three steps. First, a complaint counts as a sound complaint only if its description contains one of 18 sound words, knock through noise, after idioms such as knocked unconscious, clicks off at the pump, popped up on the screen and bells and whistles are removed. Second, a sound complaint is marked explicitly no warning if it says something like no warning lights, no check engine light, no codes or the light never came on. Third, of the rest, it is marked warning mentioned if it names any light, lamp, indicator, message, code, chime, alert or limp mode. Everything left mentions none. Reading 40 random matches in full, 39 described a real sound, and the warning group was wrong in 2 of 40, once in each direction.

Does Pulscar's study mean dashboard warning lights do not work? No, and the data cannot show that. A warning light that came on but went unmentioned is counted here as not reported, which is why the 6,184 complaints that explicitly say no light appeared are shown separately rather than folded into the headline. What the file does show is narrower and still useful: when owners describe a problem they could hear, most of them describe it without a warning. That matches how the systems are built. Federal on-board diagnostic rules are written around emission-related malfunctions, and a worn ball joint, a loose heat shield or a failing wheel bearing has no reason to set an emissions code. A noise is often the first signal a driver gets, and sometimes the only one before a tow. Treat a new sound as information, not as something to wait out until a light confirms it.

Key findings at a glance

What we counted, and what we did not

The source. NHTSA publishes its complaint database as flat files. We used the three that cover complaints received from 2015 onward, COMPLAINTS_RECEIVED_2015-2019, COMPLAINTS_RECEIVED_2020-2024 and COMPLAINTS_RECEIVED_2025-2026: 1,100,893 rows, downloaded on 13 September 2026. A complaint that names several components appears once per component, so we de-duplicated by ODINO, NHTSA's reference number, and kept product type V, which the data dictionary defines as vehicle rather than tire, equipment or child restraint. That left 777,040 complaints, received from 1 January 2015 to 8 September 2026.

The text. Every judgment is made on field 20, CDESCR, the description of the complaint. It is either the owner's account in their own words or NHTSA's standard summary of a phone call, which begins "The contact owns a" and routinely records whether a warning light was illuminated. The component category is used only to break results down afterwards, never to decide whether a complaint describes a sound.

The sound words. Eighteen whole-word patterns: knock, ticking, whine, grind, clunk, rattle, squeal, squeak, click, humming, growl, whistle, roar, screech, thump, bang, pop and noise. Before matching we strip phrases that use those words for something that is not a mechanical sound: knocked unconscious, knocked the mirror off, the pump clicks off, a screeching halt, a message popped up, bells and whistles, and no funny noises, which describes the absence of a sound. 64,831 complaints, 8.3 percent of the file, contain at least one sound word after that step.

The three groups. Each sound complaint lands in exactly one:

The explicit test runs first, so a complaint that says "no warning lights came on" is never miscounted as mentioning a light because it contains the word light. The warning-mentioned group is deliberately generous: a complaint that names the check engine light in a sentence about a different visit still counts. If anything, that pushes the quiet share down, not up.

The audits. Rules on free text make mistakes, so we read samples in full. Of 25 random complaints in the explicit group, all 25 were genuine statements that no light, code or message appeared. Of 40 random sound matches read on the final rules, 39 described an actual sound; the miss was a suspension coil that "popped off" the car. In the same 40 the warning group was wrong twice: an owner who wrote "NO WARNINGS Signs of Any Kind" landed in the warning-mentioned group because of the phrasing, and an owner who described a "Service ESC" message after a brake failure landed in the explicit group because the same complaint elsewhere says there were no lights. One error in each direction. The engine-failure and shop-could-not-find rules used below were audited the same way, 15 random matches each, and all 30 said what the rule says they say.

What we did not do. We did not rank models or brands, because complaint counts track how many of a car were sold and how motivated its owners are, not how reliable it is. We did not compute any risk of failure. We did not treat any complaint as a verified fact.

Finding 1: nearly three in four noise complaints report no warning

The headline is 73.6 percent: 47,707 of the 64,831 sound complaints report no warning light, code or message. 41,523 simply never mention one. 6,184 go out of their way to say there was none.

That second group matters more than its size suggests. People do not usually list things that did not happen. When someone filing a federal complaint stops to write "No warning lights", they expected one and it did not come. NHTSA's own phone summaries do the same: "There was no warning light illuminated" appears again and again in the standard contact narrative, recorded as a routine detail of the call.

Five owners, in their own words, from complaints received in the last two years:

• ODI 11693740, a 2015 Kia Forte, filed 16 October 2025: "Vehicle started making a ticking noise under the hood." A mechanic traced it to a loose rod bearing and said the engine had to be replaced. "There were no check engine lights that came on either."
• ODI 11736973, a 2020 GMC Sierra 1500, filed 11 May 2026: a slight knock diagnosed as a crankshaft issue grew extremely loud, with smoke from the exhaust. "No warning messages or lamps appeared."
• ODI 11641526, a 2016 Chevrolet Equinox, filed 8 February 2025: "No warning lights just loud ticking all of a sudden and then engine shaking!"
• ODI 11672851, a 2024 Toyota RAV4, filed 11 July 2025, at 13,000 miles: grinding under the middle of the car when braking. "No warning lights appeared on dash and no alerts on my mobile app."
• ODI 11681528, a 2021 Hyundai Palisade, filed 19 August 2025: brakes grinding metal on metal. "There were no lights, warning or messages to the system."

These are owner accounts, not findings. Each is included because it states what this page measures, not because the named model is worse than any other.

Finding 2: one in six owners who say no light came on describe a failed engine

A fair objection to the headline is that most noises are trivial. A squeaking seat or wind noise at a door will never light a lamp, and should not. The data agrees, for one group and not the other. Group sizes: no warning mentioned 41,523; said outright that none appeared 6,184; warning mentioned 17,124; all complaints 777,040.

The complaints that never mention a warning are mostly the minor ones. Only 3.0 percent record a tow and 5.8 percent describe an engine failure, both at or below the file average.

The owners who say outright that nothing came on are the opposite. 1,063 of them, 17.2 percent, describe an engine that failed, seized, blew or had to be replaced, nearly five times the 3.6 percent across every complaint in the file. 9.5 percent record a tow. For the knock specifically the number is stark: of 1,228 complaints in which an owner heard a knock and says no light came on, 526, or 42.8 percent, describe a failed engine. For ticking it is 98 of 373, 26.3 percent.

Read the direction carefully. This is not evidence that a missing light causes engine failures. The likelier story runs the other way: people write "no warning light" precisely when something serious and sudden happened and they are explaining why they did not stop sooner. What the table does establish is that the silent failures are not all small. More than a thousand owners in this file heard a problem, say nothing on the dashboard warned them, and describe losing an engine.

Warnings, when they come, often come late. The warning-mentioned group has the highest stall and loss-of-power share, 24.2 percent. In the samples we read, the light often appears in the same sentence as the failure itself, a flashing check engine light as the power goes, a reduced power message on the highway, rather than weeks before it.

Crash flags run lower, not higher. Sound complaints carry crash flags less often than the file as a whole, 1.6 to 2.0 percent against 4.4, most likely because a large part of the file is crash reports about air bags, seat belts and driver assistance that never mention a sound.

Finding 3: the quietest problems are the ones you can feel under you

Break the sound complaints down by the component system NHTSA filed them under. A complaint can carry more than one system, so it is counted once in each.

Suspension, wheels and structure are nearly silent. These are the parts a driver hears and feels first, a clunk over bumps, a grind in a turn, a rattle from a door, and most of them have no sensor watching. A 2020 Hyundai Palisade owner (ODI 11661447, May 2025) describes a rear shock that "started exhibiting knocking noises" at about 85,000 miles and got more frequent, confirmed by the dealer: "There were/are no warning lamps." A 2019 Lincoln MKC owner (ODI 11608594, August 2024) heard "a loud woomp woomp noise near the wheel", a wheel bearing that had to be replaced: "There were NO warning lamps or messages prior to the failure."

Steering is where the quiet gets dangerous. 80.9 percent of steering sound complaints report no warning. A 2016 Ford F-150 owner (ODI 11596870, June 2024) writes that after a right turn the "steering wheel started whining then a loud popping noise then steering wheel starts jerking back and forth then locks on me without any warning lights coming on." A steering fault light appeared only after the truck was switched off and on again. A 2016 Ford Explorer owner (ODI 11565556, January 2024) heard a "Grinding/knocking noise when driving and turning" and was told the steering gear needed replacing: "No warning lights."

Transmissions whine first and warn later, if at all. Power train complaints report no warning 73.7 percent of the time. A 2020 Toyota Highlander owner (ODI 11761227, filed 1 September 2026) says the transmission "started making a whining noise around 69,000 miles"; the dealer confirmed it and replaced the transmission. "There were no warning lights or messages." A 2020 Ford Explorer owner (ODI 11709548, January 2026), on a second transmission, describes a grinding noise and loss of acceleration while merging: "No warning lamps. No codes stored."

Brakes wear audibly, by design. Service brakes sit at 68.6 percent. Many pads carry a metal wear indicator whose whole job is to squeal, so the sound is the warning. When it turns to grinding, the stage has passed: a 2022 Nissan Pathfinder owner (ODI 11690371, September 2025) found the rear brakes failing "with no indicator on the dashboard that they needed service. Basically we just started hearing a grinding noise."

The engine warns most often, and still usually does not. Engine complaints report no warning 62.4 percent of the time, and they have the highest explicit share on the table, 12.4 percent, 2,591 complaints in which the owner expected a light and says none came. Engines have oil pressure switches, knock sensors and misfire monitoring, and plenty of engine failures do set a light, often at the end. But a bearing that has begun to knock or a timing chain that has begun to rattle can make its noise for weeks before anything electronic crosses a threshold.

Air bags are the control group. Air bag systems carry a dedicated warning lamp as part of their own self-check, and they have the lowest quiet share, 50.9 percent. Electronic stability control, another self-monitoring system, reports no warning in only 44.4 percent of its 1,229 sound complaints when broken out at the sub-system level. Where a system is built to warn, owners report warnings. Where it is not, they report sounds.

Finding 4: the newest cars are the quietest

It is tempting to assume that newer cars, with more sensors and more screens, warn their drivers more. The complaint file does not support that. Take each sound complaint's model year and the year NHTSA received it, and group by the car's age at the time:

Cars two years old or newer are the quietest group in the file. The same pattern holds if you look only at recent complaints. Among sound complaints received from 2024 to September 2026:

We cannot see from the file why, and we will not pretend to. One plausible reading is that a young car's sounds are more often build and wear issues, rattles, squeaks, early brake and suspension noise, that no sensor covers, while an older car's sounds are more often the late stage of an engine or transmission problem that eventually trips something. Either way, the practical point stands: a new car is not a reason to trust the dashboard over your ears.

Over time, the quiet share has drifted down anyway. By the year NHTSA received the complaint:

The share sat between 76 and 79 percent until 2020, dropped to 71.6 percent in 2021 and has held near 68 to 70 percent since, a fall of about ten percentage points. Put next to the age tables, that fall cannot simply be newer cars warning more; the fleet filing complaints, and the kinds of problems it files about, may have changed over the same years. In 2026 so far, 3,694 of 5,439 sound complaints still report no warning.

Finding 5: electric cars are no quieter, or louder

Electric cars have no engine to knock, and fewer moving parts, so it is reasonable to expect their complaints to look different. On this measure they do not. NHTSA's fuel-type field is blank for 95 percent of sound complaints, so we classified vehicles by model name, a heuristic described in the method file:

Electric cars land within about one point of everything else. What changes is the source of the sound: suspension, drive units and underbody parts rather than engines. A 2023 Tesla Model Y owner (ODI 11761552, filed 2 September 2026) found a "Front lateral control arm bolt spontaneously loose and about to fall out causing front tire noise" at 33,000 miles: "No warning lights." A 2012 Tesla Model S owner (ODI 11422548, 2021) heard "an abnormal humming sound" from the drive unit, which the dealer said needed replacing, with no warning light illuminated.

Hybrids report a warning more often than either, 66.7 percent quiet. With 1,599 complaints and a name-based classification, treat that as a lead rather than a result.

Finding 6: noise complaints are half again as likely to say the shop found nothing

A sound is hard to hand to a mechanic. It comes and goes with temperature, speed and load, and it often refuses to perform on a test drive. The complaint file shows the cost of that.

4.8 percent of sound complaints, 3,108 of them, say a dealer or shop could not duplicate the problem, could not find anything wrong, or called it normal. For complaints that do not describe a sound, it is 3.2 percent. Sound complaints are about 1.5 times as likely to carry that sentence.

The owners describe the same loop. A 2018 GMC Yukon owner (ODI 11244690, 2019), in capitals in the original: "My front end makes noises when I turn wheel or stop. I took it two the CarMax service department and they couldn't find anything wrong. But it's still making the noise." A 2015 GMC Yukon Denali XL owner (ODI 11005601, 2017) had a steering column bearing replaced for a clicking noise, and when the click returned "the dealer says they are unable to duplicate the noise." In the Pathfinder complaint above, the dealer called the grinding brakes normal for the vehicle.

This is the one finding on the page that points straight at something an owner can do. A recording made the moment the sound happens is evidence a technician can hear even when the car stays silent on the test drive.

Finding 7: sound by sound

The same split for each of the 18 sound words, with the tow field. A complaint that uses two words counts under both.

Knock is the outlier in both directions. It is the sound most often reported alongside a warning, 61.1 percent quiet against 74 to 83 percent for most others, and among the most often towed, 7.6 percent, second only to roar, whose 376 complaints are too few to lean on. That fits what a knock usually is: an engine problem, in the system with the most sensors, often reported at the moment the engine gives up. It also means 61.1 percent of knock complaints report no warning. Waiting for the light before taking a knock seriously means waiting on the smaller group. The engine knocking guide covers what to check first.

Squeaks and clunks are the quietest, and among the least often towed. They are what they sound like: suspension, brakes and body parts that worsen slowly and rarely strand anyone. The complete noise guide covers what each one usually is.

When in a car's life each sound shows up

About half of all complaints record the odometer at failure, field 18, MILES. Taking sound complaints with a mileage between 100 and 400,000, the median by sound word forms a ladder:

Brake and suspension squeals arrive first, around 50,000 miles; grinds and clunks in the 60,000s; whines, ticks and knocks, the drivetrain and engine sounds, last. The spread inside each word is wide and mileage is self-reported, so read this as the order in which sounds tend to appear, not as a schedule. The ticking, grinding, rattling and clicking guides go through what each one usually is.

What this study does not show

It does not show that warning lights failed. A complaint that never mentions a light is counted as not reporting one. An owner can forget, can not think it relevant, or can describe the light in words the rules do not catch. The explicit group, 6,184 complaints, is the only part of the headline in which the owner positively states that nothing came on.

It does not show that noises cause failures, or that a warning would have prevented one. Every number here is co-occurrence inside a single report.

It does not rank vehicles. Complaint counts follow sales volume, ownership age and how likely a given group of owners is to file, and nothing in the file corrects for that. The owners quoted on this page are there because of what they wrote, not because of the badge.

It does not cover every sound. Owners also write clack, clang, hiss, buzz and chirp. We kept a fixed list of 18 words so the rules stay auditable; adding words would change the totals, and the published method lets anyone test by how much.

Some classifications are heuristics. Engine failure, stalls and shop-could-not-find are text rules audited on samples, not codes in NHTSA's data. The electric and hybrid split is based on model names because the fuel-type field is mostly blank.

It is not verified data. Complaints are submitted by the public. NHTSA describes the database as a signal used with other data to identify safety issues that may warrant investigation, not as a record of confirmed defects.

What to do when your car makes a new noise

Do not wait for the dashboard to agree with your ears. In nearly three quarters of the complaints owners filed about a sound, no warning is reported, and the owners who say outright that no light came on are the ones most likely to describe a failed engine or a tow.

Write down when it happens. Cold or warm, turning or straight, braking or accelerating, which speed. That is what a technician needs, and it is exactly what a free-text complaint captures and a dashboard light does not.

Record it. Finding 6 is the reason. A 30-second recording made the moment you hear the sound is evidence that survives a quiet test drive. If a knock and a low oil level show up together, the low oil pressure guide explains why that is a stop-now combination.

Free, no card, no subscription. The report comes by email.

Download the data

• Sound complaints by NHTSA component system (CSV)
• Sound complaints by year received (CSV)
• Sound complaints by sound word, with tow counts (CSV)
• Engine failure, stalls, shop could not find, tows, crash and fire by group (CSV)
• Sound complaints by vehicle age (CSV)
• Recent complaints by model year (CSV)
• Electric, hybrid and other, from model names (CSV)
• Shop could not find it, by system (CSV)
• Exact matching rules and audit notes (TXT)

Licence. The tables and rules are published under CC BY 4.0. Please credit Pulscar analysis of NHTSA complaint data and link to this page.

For reporters. Every figure can be re-run from NHTSA's public files with the rules above. If you need a cut we did not publish, a single make, a system, a model-year band, a state, or the complaint numbers behind a table row, email [email protected] and we will send the counts and the method used to produce them.

Related reading

Our earlier, narrower study counted which noise words appear in complaints for 26 popular vehicles and which also carry crash or fire flags. For what each sound usually costs to fix, see car noises ranked by repair cost; for the light that does come on, check engine light on.

Sources and documents

Primary sources only. Every complaint quoted is retrievable by its ODI number from NHTSA's public complaints API.

1. NHTSA Office of Defects Investigation complaint flat files — COMPLAINTS_RECEIVED_2015-2019, COMPLAINTS_RECEIVED_2020-2024 and COMPLAINTS_RECEIVED_2025-2026, downloaded 13 September 2026; 1,100,893 rows, 777,040 distinct vehicle complaints by ODINO, received 1 January 2015 to 8 September 2026. File directory
2. ODI complaints data dictionary (CMPL.txt). Field 2 ODINO, NHTSA's internal reference number; field 6 YEARTXT, model year; field 12 COMPDESC, component description; field 17 LDATE, date complaint received; field 18 MILES, mileage at failure; field 20 CDESCR, description of the complaint; field 30 FUEL_TYPE; field 46 PROD_TYPE, with V for vehicle; field 49 VEHICLES_TOWED_YN, whether the vehicle was towed. Data dictionary
3. Data.gov catalogue entry for the ODI complaints dataset, describing complaint information as used with other data sources to identify safety issues that warrant investigation. Catalogue entry
4. 40 CFR 86.1806-17, the federal on-board diagnostics requirement for light-duty vehicles, written around the detection of emission-related malfunctions. eCFR
5. Quoted complaints — ODI 11751681, 11693740, 11736973, 11641526, 11672851, 11681528, 11661447, 11608594, 11596870, 11565556, 11761227, 11709548, 11690371, 11761552, 11422548, 11244690 and 11005601. Example query
6. Pulscar study files — nine aggregate tables and the exact rules used on this page. Method and rules

How these numbers were built: one script over NHTSA's public complaint flat files, run on 13 September 2026; the sample, the sound words, the idioms removed, the three-group classification, the outcome rules, the audits and the limits are all described above and in the published method file. No model or brand ranking and no failure risk is computed anywhere on this page.

Prices verified: not applicable, this study quotes no repair prices · Technical documents verified: 13 September 2026 · Data refreshed: 13 September 2026

Re-ran the rules and got a different number? Email [email protected] with your counts and the file dates you used.