Get a Full Vehicle History Report

Do NHTSA Complaint Numbers Mean a Used Car Is Unreliable? We Audited Every Complaint Category for Recall Paperwork

· Zilocar Editorial

TL;DR: No — not by themselves. In an audit of all 776,375 distinct vehicle complaints in NHTSA's public ODI database (retrieved 2026-09-14), 152,796 — 19.7%, about one in five — reference an existing recall rather than report an independent fault. That share reaches 63.1% in fuel-system complaints and at least 47.6% in airbags. Removing recall correspondence flips the most-complained-about model in 3 of 7 categories, so a raw complaint count can measure recall reach, not reliability.

Users who trust ZILOCAR
Trusted by thousands of users all over the globe

Key facts

  • Dataset: NHTSA ODI COMPLAINTS_RECEIVED period files (2015-2019, 2020-2024, 2025-2026), retrieved and processed 2026-09-14. Newest filing in the corpus is dated 2026-09-08.
  • N = 776,375 distinct vehicle complaints (PROD_TYPE=V, deduplicated on complaint ID) drawn from 1,086,951 component rows.
  • Corpus-wide, 19.7% (152,796) of complaints reference an existing recall — roughly one in five, not "half."
  • Most contaminated of the seven categories audited: fuel system (gasoline), 63.1% recall-referencing. Airbags, 47.6%. Least contaminated: crash-injury, 10.4%. Fuel's lead is not robust to how the brake category is defined — see Method.
  • One recall dominates fuel: Honda's in-tank fuel-pump campaign 23V858 (2,490,460 vehicles) supplies 17.5% of the entire fuel-system bucket — the most concentrated single campaign in any category.
  • 5 of 7 categories flip at least one "most-complained-about model" ranking once recall correspondence is removed. Electrical and crash-injury hold on both.
  • News peg: Ford 26V578 (223,472 MY2023-2027 F-150, fuel-tank straps; owner letters expected 2026-09-21) and Jeep/Chrysler 26V562 (328,381 Grand Cherokee, rear coil springs; interim letters expected 2026-09-29) are pouring fresh recall correspondence into the same complaint stream.
  • All contamination figures are conservative lower bounds — the text-matching classifier misses some recall references.

By the numbers

  • 776,375 distinct vehicle complaints audited (1,086,951 component rows before de-duplication).
  • 152,796 / 19.7% reference an existing recall corpus-wide.
  • 63.1% — fuel system (gasoline), the most contaminated of the seven categories as defined here (joint-highest with hydraulic brakes once NHTSA's two brake strings are scored separately; see Method).
  • 47.6% — airbags (at least; the true share is higher, see method note).
  • 17.5% — share of the entire fuel bucket supplied by one campaign, Honda 23V858.
  • 3 of 7 categories change volume leader after recall complaints are stripped; 4 of 7 change share leader; 5 of 7 flip at least one.

Based on 776,375 distinct NHTSA ODI vehicle complaints across seven COMPDESC categories, computed by Zilocar from the three COMPLAINTS_RECEIVED period files (max filing dates 20191231 / 20241231 / 20260908), retrieved 2026-09-14. Frozen results CSV linked in Sources.

Do NHTSA complaint numbers mean a used car is unreliable?

Not on their own. A raw NHTSA complaint count blends two very different things: owners reporting a fault they experienced, and owners logging a note about a recall — a letter that arrived, a remedy that was delayed, a part on backorder. Across the whole database, 19.7% of complaints reference an existing recall rather than an independent fault. In some categories it is a clear majority. So a scary count can measure a car's recall reach and its sales popularity as much as its reliability.

Two structural caveats compound this. Complaints are counts, not rates — there is no sales denominator in the data, so a common nameplate draws more complaints than a rare one regardless of quality. And large recalls generate a burst of correspondence that lands in the same totals used to build "most-complained-about" lists. This is the same confound we found when we asked whether owner complaints predict a car recall: the recall and the complaints feed each other.

Are complaints about recalls counted in a car's NHTSA complaint total?

Yes. NHTSA's Office of Defects Investigation records what owners submit in free text, and owners routinely use the complaint form to document recall experiences — citing a campaign number, describing a notification letter, or reporting that the remedy part is unavailable. Those entries are not separated out; they sit in the same COMPDESC categories and the same model totals that third-party "most-complained-about" rankings count. That is why a single mega-recall can visibly move a model up a leaderboard.

Which NHTSA complaint categories are most inflated by recall paperwork?

Fuel system (gasoline) is the most contaminated category at 63.1% recall-referencing, followed by airbags at 47.6%. The remaining categories fall well below half: electrical 19.3%, brakes 18.2%, suspension 16.8%, engine 14.6%, and crash-injury 10.4% (least contaminated). We tested the "most contaminated" and "biggest single campaign" claims against all seven buckets. Fuel wins the single-campaign claim outright. On contamination it wins as these seven categories are defined, but not unconditionally: our "service brakes" bucket is a prefix match that merges NHTSA's two separate brake strings, and the smaller of them, SERVICE BRAKES, HYDRAULIC (7,420 complaints), runs about 63% on its own — level with fuel. Read fuel and hydraulic brakes as joint-highest rather than crowning one — Honda's 23V858 fuel-pump recall alone accounts for 17.5% of the fuel category, while every other category's top campaign sits below 2.6%.

Category (COMPDESC bucket)ComplaintsRecall-referencingContam. rateTop campaign (share of bucket)Volume leader as-filedVolume leader after stripVolume flips?
Fuel system, gasoline7,3044,60863.1%Honda 23V858 (17.5%)Honda CR-V (429)Ford Escape (93)YES
Air bags97,89446,59047.6%Ford 16V384 Takata (2.6%)Toyota Corolla (3,880)Ford Fusion (1,297)YES
Electrical system137,51226,52219.3%Chrysler 14V373 (1.8%)Jeep Grand Cherokee (5,549)Jeep Grand Cherokee (2,564)no
Service brakes70,26312,75918.2%Kia 23V652 (0.9%)Ram 2500 (2,996)Chevrolet Silverado 1500 (1,998)YES
Suspension30,9105,20616.8%Toyota 14V604 (1.1%)Ford F-250 SD (1,806)Ford F-250 SD (1,700)no
Engine157,14822,92514.6%Ford 23V905 (0.6%)Ford Escape (7,133)Ford Escape (6,225)no
Crash / injury (severity flag)40,2594,19510.4%Chrysler 15V313 Takata (0.2%)Jeep Grand Cherokee (960)Jeep Grand Cherokee (812)no
Corpus (all buckets)776,375152,79619.7%

Only the fuel bucket clearly exceeds half. Airbag sits just under half here and is higher in reality — our classifier misses brand-name-only references ("Takata" with no "recall" token), so a broader any-mention variant reads airbag near two-thirds. The corpus-wide "half is recall paperwork" framing is therefore true only for specific categories, not the database as a whole.

Which "most-complained-about model" rankings flip when recall mail is removed?

Five of the seven categories change at least one leaderboard once recall-referencing complaints are stripped out — direct evidence that the raw counts partly rank recall size rather than reliability. On volume, the top model changes in fuel (Honda CR-V to Ford Escape), airbags (Toyota Corolla to Ford Fusion), and brakes (Ram 2500 to Chevrolet Silverado 1500) — the same reversal we saw ranking used cars with the most brake failure complaints. On share of a model's own complaints, the leader changes in airbags, fuel, suspension, and engine.

Two categories hold on both measures. Our ranking of used cars with the most electrical problems is unchanged (Jeep Grand Cherokee on volume, Kia EV6 on share), and so is crash-injury (Jeep Grand Cherokee on volume, Tesla Model Y on share). The list of used cars with the most engine failure complaints also keeps its volume leader (Ford Escape) even though its share leader flips. Where a ranking survives the strip, the signal is more likely to be about the car; where it flips, the original ranking was measuring the recall.

How do I tell a real defect complaint from a recall-correspondence complaint on NHTSA?

Read the narrative, not the count. A recall-correspondence complaint typically names a campaign number (like 23V858 or 26V578), describes a letter or notification, mentions a "remedy," or reports a part on backorder — often without describing a fault the owner actually experienced. A defect complaint describes a symptom: a leak, a stall, a warning light, a noise, a failure to start.

Our classifier applies exactly that logic. A complaint is counted as recall-referencing if its free text matches a campaign-number pattern (\d{2}[VES]\d{3}), uses safety-recall language (the word "recall" near notice/letter/campaign/remedy/repair/issued/received/subject-to/part-of), says "remedy," or reports a part as unavailable or backordered. The reported rates are conservative lower bounds: the pattern misses references phrased outside it (for example "recall has been issued") and brand-name-only mentions, which biases contamination downward — most for airbags. Treat the loose contamination share as the primary metric; a stricter "recall note with no fault described" cut exists in the data but over- and under-counts on component nouns and is illustrative only.

Does the Ford F-150 or Jeep Grand Cherokee recall mean I shouldn't buy?

Neither recall is, by itself, a reason to walk away — and both illustrate why this week's rankings need reading with care. Ford's 26V578 covers 223,472 MY2023-2027 F-150s whose fuel tank may be improperly secured; the free remedy is to inspect and replace the fuel-tank straps, owner letters are expected 2026-09-21, and VINs became searchable on NHTSA.gov on 2026-09-14. It is not a do-not-drive or park-outside order. Jeep/Chrysler's 26V562 covers 328,381 Grand Cherokee, Grand Cherokee L and 4XE units for rear coil springs that may be incorrectly installed; the free remedy replaces the rear lower isolators, interim letters are expected 2026-09-29, and it supersedes earlier campaigns 26V051 and 23V413 (vehicles fixed under those need the new remedy).

Both recalls will generate a wave of owner correspondence over the coming weeks — exactly the recall paperwork that inflates fuel and suspension complaint counts. What matters for a specific truck or SUV is whether that VIN's recall is open and whether it has been repaired, not the model's aggregate complaint tally.

What a VIN check can and can't tell you here

A Zilocar VIN check is a per-car complement to this model-level analysis, and its honest role here is specific. It can confirm whether a specific vehicle sits inside an open safety recall and how many are attached — the same presence signal as NHTSA's free VIN look-up — and it adds accident and damage records (including airbag-deployment status), an odometer/rollback check, theft (NICB) and junk/salvage auction records, ownership history, and sales-listing history. It does not map a VIN to complaint counts and does not filter recall correspondence out of a model's complaint total; that de-confounding is what this dataset does at the model level, not something any per-car report does.

QuestionVIN checkWhere to confirm
Is this car under an open recall (e.g. 26V578, 26V562)?Yes — recall presence/countNHTSA VIN look-up; Zilocar
Was the recall actually repaired/flashed?NoFranchise dealer or manufacturer owner portal
Is this model heavily represented in NHTSA complaints?Not per VINThis model-level analysis
How much of that count is recall paperwork vs. faults?NoThis analysis
Is there an open NHTSA investigation (PE/EA)?NoNHTSA.gov
Accident, airbag deployment, salvage-auction, odometer, ownership, listing history?YesZilocar

Because a recall count and a complaint count both climb after a big recall, remedy confirmation is the piece that actually protects a buyer — and that is exactly what a VIN check cannot show. For why an unrepaired recall persists on a used car, see does a completed recall mean a used car is fixed. Confirm remedy status with a franchise dealer using the VIN.

A Zilocar VIN check screens a specific car for recall presence and pulls its accident, airbag-deployment, salvage-auction, odometer, theft, ownership and sales-listing history, plus specs, NHTSA and IIHS safety ratings, and a market valuation. It does not confirm a recall was remedied, track NHTSA investigations, show firmware status, or classify the legal title brand — check those with NHTSA and a dealer.

Frequently asked questions

Method

Source files. NHTSA Office of Defects Investigation consumer-complaint period files, not the combined download: COMPLAINTS_RECEIVED_2015-2019.zip (81,302,701 bytes), COMPLAINTS_RECEIVED_2020-2024.zip (75,873,826 bytes) and COMPLAINTS_RECEIVED_2025-2026.zip (38,983,619 bytes), all retrieved and processed 2026-09-14. Tab-delimited, 51 fields, no header row, layout per CMPL.txt.

Freshness was asserted before anything was computed. · max(LDATE) per file is 2019-12-31 / 2024-12-31 / 2026-09-08, so the corpus's newest filing is six days before this run. The combined FLAT_CMPL.zip was deliberately not used: on 2026-09-13 it served 56,997 rows whose newest filing was 1996-10-10 — 3.2MB against 371MB three days earlier — with the archive's CRC intact and an unremarkable Last-Modified, and on 2026-09-08 it served records stopping at 2018-07-14. Two silent regressions in six days is the normal behaviour of that endpoint, so the period files are the default here. Neither file size nor timestamp distinguishes a good copy from a broken one; only the newest filing date does.

Universe. Keep PROD_TYPE = 'V', dropping equipment, tyre and child-seat rows, and LDATE >= 2015-01-01. Rows are one per cited component, so 1,086,951 component rows were deduplicated on ODINO to N = 776,375 distinct complaints; no duplicate CMPLID appears across the three files. Every figure on this page counts complaints — never rows, vehicles, owners or confirmed faults.

Categories. Six are COMPDESC prefix matches (AIR BAGS, FUEL SYSTEM, GASOLINE, ELECTRICAL SYSTEM, SERVICE BRAKES, SUSPENSION, ENGINE). The seventh, crash-injury, is a severity cut (CRASH='Y' or INJURED>0 or DEATHS>0) across any component. Each definition is stated verbatim in the frozen CSV's compdesc_definition column.

The classifier, which is the load-bearing step. A complaint is recall-referencing when its free-text CDESCR cites a campaign number (\d{2}[VES]\d{3}), uses safety-recall language (recall adjacent to notice / notification / letter / campaign / repair / remedy / part / issued / received / subject to / part of), says remedy, or reports a part unavailable or backordered. Strict recall-only additionally requires that the text describe no fault symptom.

⚠️ The strict figure is not comparable with our 2026-09-13 fuel-system study, and the gap is a definition rather than a correction. That article used a fuel-specific fault screen (fuel leak, pump, tank, injector, rail, filler, gauge, canister) and classified 3,047 fuel complaints — 41.7% — as pure recall administration. The category-agnostic screen used here counts much more text as fault-describing, so the same bucket reads 121, or 1.7%. Both are internally consistent under their own definitions and neither supersedes the other, but they must not be quoted side by side. The loose measure is the one that is comparable across the two pages: 58.9% there, 63.1% here, both majorities, both reversing the same two leaderboards.

⚠️ The "most contaminated category" ranking is definition-dependent at the very top. · SERVICE BRAKES above is a prefix match that merges NHTSA's two separate top-level brake strings — SERVICE BRAKES (62,744 complaints) and SERVICE BRAKES, HYDRAULIC (7,420). Scored on its own, the hydraulic string runs about 63% recall-referencing, statistically level with fuel's 63.1%. So "fuel is the most contaminated category" holds under the seven categories as defined here and does not survive splitting that prefix into its two real NHTSA strings. Fuel and hydraulic brakes should be read as joint-highest. This is the same species of trap as NHTSA's free-text MODELTXT splitting one nameplate across several strings, one level up in the hierarchy.

Independent re-derivation. The whole corpus was recomputed a second time from separately downloaded copies of the three archives, using a classifier written from scratch rather than reused. Bucket totals reproduce to within 0.1% — fuel 7,308 against 7,304, airbags 98,027 against 97,894, electrical 137,631 against 137,512, suspension 30,928 against 30,910 — the distinct-complaint universe to within 664 of 776,375, and both headline reversals reproduce identically: airbags Toyota Corolla to Ford Fusion, fuel Honda CR-V to Ford Escape. The independent classifier read every bucket 2 to 6 points lower (fuel 61.2%, airbags 41.0%), which is the expected spread between two hand-written text screens and is why the contamination rates here are presented as points in a range rather than exact values.

Known limits. (1) Complaints are consumer-reported and unverified; NHTSA does not adjudicate them. (2) There is no denominator — nothing here is a rate, and a popular nameplate draws more complaints regardless of quality. (3) The classifier is approximate in both directions: it misses brand-name-only references (a "Takata" mention carrying no recall token), which is why airbags are stated as at least 47.6%, and it flags genuine fault reports that mention a recall in passing. (4) A COMPDESC count is an upper bound on the component independently of the recall issue. (5) FIRE, INJURED and DEATHS are owner-entered floors, not measurements. (6) Model-level only — nothing here describes an individual vehicle. (7) NHTSA republishes these files continuously, so a re-pull will not reproduce these counts exactly; the frozen CSV is the reproducible artefact.

Sources

Last verified: 2026-09-14