America's Least-Fixed Recalls: The Mega-Recalls Leaving Millions of Cars Driving With Open Safety Defects

Read the 148.9 million correctly: it counts recall-vehicle pairs, not distinct cars. A vehicle under three open recalls appears three times. Across the dataset there are 450.3 million obligations against a US light-vehicle fleet of roughly 290 million, about 1.55 obligations per vehicle, so the number of individual affected cars is materially smaller. We publish the denominator so the ratio, not the headline, is the citable figure.

Converted to distinct cars, 148.9 million obligations divided by 1.55 is roughly 96 million vehicles, about a third of the fleet. That sits in the same territory as the widely cited industry estimate of roughly one in four vehicles carrying an open recall, and we would rather show that reconciliation than let the headline imply we are claiming double it. The fleet figure used here is the roughly 290 million light vehicles in operation reported by S&P Global Mobility; Ford's build volume is a Zilocar estimate from its own US sales across the covered model years.

Key findings

1. Ford's 1.45-million-vehicle rearview-camera recall (25V695000, Oct 2025) is 1.09% fixed — about 1.43 million cars still driving with the defect.
2. Ford carries an estimated 18.2 million unrepaired recall obligations from recalls since 2020 — more than GM (5.9M), Honda (5.3M) and Stellantis (5.3M) combined. Applying Ford's own 1.26 obligations-per-vehicle overlap, that is roughly 14.4 million distinct vehicles. That is the largest unrepaired pool, not the worst rate: BMW (53.8%), Nissan (58.1%), Kia and Hyundai (both 60.5%) all fix a smaller share than Ford's 61.0%.
3. 96 mega-recalls since 2020 (100k-plus vehicles each) sit under 50% completion, covering 46.0 million vehicles.
4. Across all 10,494 completion-tracked recalls, an estimated 148.9 million recall obligations remain unrepaired out of 450.3 million; overall units-weighted completion is 66.9%.
5. Honda's Takata driver-side air-bag recall (20V026000) is still only 7.91% fixed six years after the 2020 remedy — about 1.33M of 1.44M inflators. It has not moved in three weeks of re-measurement.
6. Safety-critical recalls are fixed slower than convenience ones, 65.2% vs 68.6% — but that is an air-bag composition effect, not a danger effect. Air bags alone are 60.7% across 110.8M vehicles, and with them removed the safety bucket rises to 69.3%, marginally ahead of convenience recalls.
7. Fix rates can move fast, which is why an as-of date matters: Ford's 4.38-million-vehicle F-150 trailer-brake recall (26V104000) went from 16.12% fixed on July 22 to 65.65% on August 12, cutting its unfixed pool from 3.68 million trucks to 1.51 million in three weeks.

By the numbers

• 1.09% fixed — Ford's worst recall of 500,000-plus vehicles (25V695000, 1,448,655 vehicles), about 1.43 million still unrepaired. The lowest completion rate of any recent 1M-plus recall in the data.
• 45,972,468 vehicles — the pool sitting across 96 recalls (since 2020, 100k-plus each) that are less than half-fixed.
• ~148.9 million unrepaired obligations out of 450,259,447 total across all 10,494 completion-tracked recalls, at 66.9% overall units-weighted completion.

Every figure below is from a Zilocar analysis of NHTSA "Recalls Data" (Socrata dataset 6axg-epim), data as of August 12, 2026. The exact snapshot used is published as a CSV in Sources, so these numbers stay checkable after NHTSA's live file moves on.

In the middle of the loudest recall year the U.S. auto industry has ever recorded, the number nobody is publishing is the one buyers actually care about: not how many cars got recalled, but how many are still broken.

Here is one answer. In October 2025, Ford recalled 1,448,655 vehicles because the rearview camera image could fail to display. Ten months later, that recall is 1.09% fixed. That is roughly sixteen thousand cars repaired out of nearly one and a half million. About 1.43 million are still on the road with the defect NHTSA flagged.

Only one recall of a million-plus vehicles in the entire dataset is fixed less, and it covers cars built in the 1990s. And it is not alone. When you rank recalls not by how big they are but by how few have been fixed, a national backlog comes into focus: an estimated 148.9 million recall obligations that were never discharged.

Finding 1: One Ford recall is 1.09% fixed

Ford's 25V695000 is the sharpest example of a pattern that runs through the recent data: enormous recalls, tiny fix rates. It is verified to the unit against NHTSA's own recalls API (which, notably, carries no completion figure — that data lives only in dataset 6axg-epim).

It is not even Ford's only badly lagging camera recall. A separate September 2025 campaign (25V572000) covers another 1,456,417 vehicles at 15.38% fixed. Together, two Ford rearview-camera recalls account for roughly 2.67 million vehicles still unrepaired.

Table 2 — 10 biggest recent recalls (received since 2024-01-01) and how many are still unfixed

Honest annotation: several of these are recent, so completion is still climbing — owner notification lags months. Tesla's 99.3% shows what a fixable-by-OTA software recall looks like. Between July 22 and August 12, 2026 this table moved a lot: the F-150 trailer-brake recall went from 16.12% to 65.65% fixed and Toyota's camera recall from 66.19% to 81.09%, while Ford's 25V695000 crawled from 0.70% to 1.09%. Recalls that are going to get fixed tend to move quickly; the ones that stall, stall from the start.

Finding 2: Ford carries more unrepaired recall obligations than GM, Honda and Stellantis combined

The record-recall story of 2025-2026 has been about Ford's sheer volume. The completion data adds the second half of the sentence: those recalls are also completed at a below-average rate.

Since 2020, Ford issued 442 recalls covering 46.8 million vehicles at 61.0% units-weighted completion, below the 66.9% dataset-wide average. That leaves an estimated 18.2 million unrepaired Ford recall obligations, more than General Motors, Honda and Stellantis combined. Ford's 46.8 million units sit against roughly 37 million vehicles built in the covered model years, about 1.26 obligations per vehicle, so the distinct-vehicle figure is nearer 14.4 million.

One claim we are not making: Ford does not have the worst fix rate. Four major automakers complete a smaller share of their recalls than Ford does — BMW at 53.8%, Nissan at 58.1%, Kia at 60.5% and Hyundai at 60.5%. Ford leads this ranking on the size of its unrepaired pool, which is a function of volume as much as diligence. An earlier version of this article called Ford's rate the lowest of any major automaker. That was wrong, and it contradicted our own Table 3.

Table 3 — Automaker ranking: est. unrepaired recall obligations (all recalls received since 2020-01-01)

Volume drives the ranking, but not entirely. BMW recalled roughly a quarter as many vehicles as Toyota yet fixed a far smaller share of them (53.8% against 73.1%), and Nissan's 58.1% is lower than Ford's. Ranked by rate rather than volume, the order at the bottom is BMW, Nissan, Kia, Hyundai, then Ford — a reminder that a smaller automaker with a stubborn fix rate still leaves millions of cars unrepaired, and that the biggest pool is not the same thing as the worst record.

Finding 3: Nearly 50 million vehicles are less than half-fixed

Zoom out from any one brand and the scale of the backlog is a national figure. Since 2020, 96 recalls of 100,000 vehicles or more sit under 50% completion, together covering 45,972,468 vehicles — a pool of nearly 46 million cars that are less than half repaired. That count was verified against a server-side aggregate of the same dataset.

Finding 4: The never-fixed classics stay stuck for years

Recency explains the low fix rates in Table 2. It cannot explain Table 1. Here the filter is strict: only recalls of 500,000-plus vehicles where the remedy has been available two or more years. These are cars whose owners have had ample time — and the fix rates are still dismal.

Honda's Takata driver-side air-bag recall (20V026000) is 7.91% fixed six years after its 2020 remedy: about 1.33 million of 1.44 million inflators that can rupture are still installed. That figure did not move at all between our July 22 and August 12 measurements. Ford's Explorer A-pillar trim recall (24V031000, 1.89 million vehicles) sits at 20.84%.

Table 1 — America's most chronically unfixed mega-recalls (500k+ vehicles, remedy available 2+ years, ranked by lowest completion)

A caution reading the top of this table: the two lowest entries are old and unusual. Mazda's 0.62% recall covers 1989-1998 vehicles, many long scrapped, and Keystone's covers RVs — their fix rates are depressed by vehicles that no longer exist and owners who cannot be reached, not only by neglect. That is exactly why Ford's recent, reachable-owner camera recall (1.09%) is the cleaner "barely fixed at all" headline. But from the Honda Takata rows down, these are ordinary passenger cars with living owners and available parts, still stuck in the teens and twenties years later.

Finding 5: Life-or-death recalls aren't fixed any faster

You might assume the scariest recalls get prioritized. They don't. Bucketing all 10,494 recalls by defect type, safety-critical recalls (air bags, steering, brakes, fuel and fire, powertrain, suspension, tires, seat belts) are fixed at 65.2% units-weighted completion — slower than convenience and electronics recalls at 68.6%.

Table 4 — Do safety-critical recalls get fixed faster? (units-weighted completion by defect type)

Air bags alone sit at just 60.7% completion across 110.8 million vehicles — 48% of the entire safety-critical bucket by units.

That single backlog drives the whole result, and we tested it rather than leaving you to. Strip air bags out and safety-critical recalls sit at 69.3%, which is statistically indistinguishable from convenience recalls at 68.6% and actually slightly higher. So the honest finding is not that danger makes owners slower. It is that one unresolved defect class, Takata inflators, is large enough on its own to drag the entire safety-critical category three points below a glitchy backup camera. The comforting assumption that dangerous recalls take care of themselves is still wrong, but for a more specific reason than "people ignore scary recalls". Anyone can reproduce this in ten minutes from the CSV linked in Sources.

Finding 6: The biggest recall of 2026 moved 49 points in three weeks, and the stalled ones didn't move at all

This finding replaces an earlier one, and the reason it changed is itself the point.

The largest live recall of the year is Ford's February 2026 F-150 campaign (26V104000): 4,381,878 trucks whose trailer lighting and brakes may not function. When we first measured it on July 22, 2026 it stood at 16.12% fixed, roughly 3.68 million pickups still carrying the defect. Re-measured on August 12, it is 65.65% fixed, with about 1.51 million left. Nearly 2.2 million trucks were repaired in three weeks.

That is what a functioning recall looks like once parts and appointments catch up, and it is the strongest argument against reading any single recent recall as scandal. It also means a completion figure without an as-of date is close to worthless.

Now hold that against the recalls in Table 1. Honda's Takata inflator recall sat at 7.91% on July 22 and sits at 7.91% today, unmoved. Ford's 25V695000 crawled from 0.70% to 1.09%. The distinction that matters is not big versus small or new versus old, it is moving versus stalled. Recalls that are going to be fixed start being fixed quickly. The ones that are still in the teens after two years are not waiting on parts.

Methodology

Source: NHTSA "Recalls Data," Socrata dataset 6axg-epim on data.transportation.gov (no auth). On 2026-08-12 I pulled a bulk CSV of all vehicle-type recalls carrying a populated completion rate via $where=recall_type='Vehicle' AND completion_rate IS NOT NULL ($limit=50000, no pagination needed): N=10,494 recalls, report-received dates 2011-05-25 to 2026-06-15 (completion rates are populated by NHTSA quarterly/annual reports from 2015-Q1 onward). Cleaning: exactly 3 artifact rows carried completion_rate above 100% (20V731000 at 133.33% on 3 units, 20V664000 at 3,380% on 5 units, 19V871000 at 108.53% on 140 units) and were capped at 100%; zero rows had potentially_affected <= 0. Per row I computed estimated_unfixed = round(potentially_affected * (1 - min(completion_rate,100)/100)); potentially_affected is the manufacturer's at-filing population estimate. Table 1 filters to potentially_affected>=500,000 and report_received_date<=2024-08-12 (remedy available 2+ years) to dodge the recency confound; Table 2 to report_received_date>=2024-01-01; Table 3 aggregates report_received_date>=2020-01-01, subsidiary names normalized to parent brand; Table 4 buckets the NHTSA component field. Headline recalls' affected counts and subject text were verified against the independent api.nhtsa.gov/recalls/campaignNumber API, and the 96-recall aggregate was cross-checked against a server-side SoQL query.

This dataset is replaced daily, and these figures are a snapshot. The article was first published on 2026-07-22 and re-measured in full on 2026-08-12; every number above is the August 12 value. Between those dates one headline recall moved 49 percentage points (see Finding 6). The exact CSV snapshot behind this version is published in Sources so the figures stay verifiable after NHTSA's live file moves on. One honest limit on that: the 2026-07-22 snapshot was not retained, so where this article compares a July value against an August value, only the August side can be reproduced from the published file. The July figures are the ones this article published on that date. If you re-run the query in Sources today you will get different numbers, and that is expected rather than a discrepancy.

What the numbers can and cannot tell you:

• The 148.9 million counts recall obligations, not distinct cars. This is the single most important caveat and the easiest to get wrong. Summing affected populations across recalls treats every recall as covering a different set of vehicles, which is false: a truck recalled three times appears three times. Ford's recalls since 2020 sum to 46.8 million units against roughly 37 million US vehicles built in the covered model years, and campaigns 25V315000 and 26V104000 recall overlapping F-150 populations. Dataset-wide there are 450.3 million obligations against a US light-vehicle fleet of roughly 290 million, about 1.55 per vehicle. The defensible statistic is the ratio — roughly one in three recall obligations is never discharged — not a headcount of cars. Anyone citing a number of individual unrepaired vehicles should use an industry estimate built from VIN-level data, not this figure.
• "Estimated unfixed" is directional, not a headcount. It is affected population multiplied by (1 − % fixed). "Affected" is the manufacturer's at-filing estimate, and "% fixed" is a snapshot from NHTSA's latest quarterly/annual reporting — not a VIN-confirmed live count of cars still needing repair.
• Completion coverage starts at 2015-Q1. Recalls before that, and any recall NHTSA has not reported a rate for, carry no completion figure and are excluded. N=10,494 is the completion-tracked universe, not every recall ever issued.
• Recent recalls look worse partly because they are recent. Owner notification and parts supply lag months, so Table 2's low percentages partly reflect timing. Table 1 removes this by requiring the remedy to have been available two-plus years.
• The very lowest outliers are inflated by scrapped cars and unreachable owners. Mazda 15V674000 (0.62%) covers 1989-1998 vehicles and Keystone 17V755000 (2.15%) covers RVs. Ford 25V695000 (1.09%, MY2015-2019) is the strongest recent barely-fixed example. Mazda edges Ford at the very bottom; it is reported honestly in Table 1, but the recent, reachable-owner Ford recall is used as the headline outlier.
• Rate and pool size are different rankings and we report both. Ford tops Table 3 on estimated unrepaired vehicles because it recalls the most vehicles. On completion rate since 2020, BMW (53.8%), Nissan (58.1%), Kia (60.5%) and Hyundai (60.5%) all sit below Ford (61.0%). Any claim that one automaker is "worst" must say worst at what.
• Brand grouping and component buckets carry residual error. Table 3 normalizes subsidiary names to parent brands, with two known exceptions we state rather than hide: the Toyota row counts Toyota Motor Engineering and Manufacturing only, excluding Gulf States Toyota and Southeast Toyota Distributors (adding them gives 93 recalls and 4,662,390 obligations, no rank change), and the Volvo row merges Volvo Car USA with Volvo Trucks and Volvo Bus, which have been separate corporations since 1999 (Volvo Cars alone is 539,202 and still holds rank 12). Residual mis-grouping is possible for smaller brands. Table 4 uses NHTSA's component field across 35 distinct values, and 4,684 recalls fell into "other/unclassified" and are shown separately rather than forced into a bin. The bucketing is a judgment call and the component column is in the published CSV, so anyone can re-cut it. Two mappings worth knowing before you quote Table 4: the safety bucket includes engine and engine cooling, and Ford 26V104000 (trailer lighting and brakes, the year's largest recall) carries the component value ELECTRICAL SYSTEM and therefore counts as convenience despite being a braking defect.
• Completion could not be triangulated against a second source. The api.nhtsa.gov recalls API used for belt-and-suspenders verification carries no completion rate; fix-rate figures exist only in dataset 6axg-epim.

What this means if you're buying used

A national completion rate is not a verdict on any one car. But it does reset the odds. With roughly one in three recall obligations never discharged, a used car you are considering may well carry recall history worth checking — and the older, higher-mileage cars that dominate the used market are exactly the ones caught in these long-tail recalls.

The honest limits matter here. A completion percentage is a national snapshot, not a per-vehicle status, and the only way to know a specific car is to check its VIN. A Zilocar VIN check surfaces whether a recall is present or outstanding on a given vehicle (recall presence and count — not the remedy or firmware status), alongside accident and airbag-deployment records, odometer and rollback discrepancies, salvage and junk auction records, theft records, ownership count and past sales-listing history. It does not confirm that a recall was actually performed — for that, run the VIN through NHTSA's free recall lookup and have a dealer verify the remedy before you buy. Used together, they turn a scary national statistic into a specific, answerable question about the car in front of you.

FAQ

How many recalled cars in the U.S. are still unfixed? Across 10,494 completion-tracked vehicle recalls in NHTSA's data, a Zilocar analysis estimates roughly 148.9 million unrepaired recall obligations out of 450.3 million total, because overall units-weighted completion is 66.9%. That counts recall-vehicle pairs, not distinct cars: a vehicle under three open recalls counts three times. The citable statistic is the ratio, roughly one in three, not a headcount. Data as of August 12, 2026.

What is the least-fixed car recall in America right now? Among recalls of 1 million-plus vehicles, Ford's October 2025 rearview-camera recall (25V695000) covers 1,448,655 vehicles and is only 1.09% fixed. An older Mazda ignition-switch recall (15V674000) is technically lower at 0.62%, but it covers 1989-1998 vehicles, many long scrapped.

Which automaker has the most unrepaired recalled vehicles? Ford, by volume: an estimated 18.2 million unrepaired Ford recall obligations from recalls since 2020, more than GM, Honda and Stellantis combined, roughly 14.4 million distinct vehicles after overlap. Ford issued 442 recalls covering 46.8 million vehicles at 61.0% units-weighted completion. But Ford does not have the worst fix rate — BMW (53.8%), Nissan (58.1%), Kia (60.5%) and Hyundai (60.5%) are all lower.

Do dangerous safety recalls get fixed faster than minor ones? Slightly slower overall, but the gap is an air-bag composition effect rather than a danger effect. Safety-critical recalls sit at 65.2% completion versus 68.6% for convenience and electronics. Air bags are 48% of the safety bucket by units and only 60.7% complete; remove them and safety-critical rises to 69.3%, marginally ahead.

Does a low national completion rate mean a specific car still has an open recall? Not necessarily. Completion rate is a national snapshot from manufacturer reporting, not a per-vehicle status. Check the VIN — a VIN check surfaces whether a recall is present or outstanding, and NHTSA's free lookup plus a dealer can confirm the remedy.

Why are some recent recalls at such low completion rates? Partly because they are recent: owner notification and parts supply lag months. Table 1 removes this by only including recalls where the remedy has been available two or more years.

Related checks

• used ford recalls to check before buying
• used honda odyssey problems to check before buying
• used car brands with the most recalls
• used ford focus problems to check before buying
• what does recall completed mean when buying a used car

Sources

• Download the exact data behind this article (CSV, 10,494 rows, 1.5 MB, snapshot taken 2026-08-12). This is the frozen file every figure above was computed from, so you can reproduce them after NHTSA's live dataset moves on. https://zilocar-news.github.io/zilocar-rss/data/nhtsa-recall-completion-2026-08-12.csv
• NHTSA "Recalls Data," Socrata dataset 6axg-epim, data.transportation.gov (primary dataset; completion rates from 2015-Q1 onward). Note this file is replaced daily, so a live query will not match the snapshot above. https://data.transportation.gov/resource/6axg-epim
• Live query used to build the snapshot (vehicle recalls with a populated completion rate). https://data.transportation.gov/resource/6axg-epim.csv?$select=nhtsa\_id,report\_received\_date,subject,manufacturer,component,potentially\_affected,completion\_rate,recall\_type&$where=recall\_type='Vehicle'%20AND%20completion\_rate%20IS%20NOT%20NULL&$limit=50000
• NHTSA Recalls API (api.nhtsa.gov/recalls/campaignNumber), used to verify affected counts and subject text to the unit; note it carries no completion rate. https://api.nhtsa.gov/recalls/campaignNumber?campaignNumber=25V695000
• NHTSA Recalls consumer lookup (check a specific VIN's open-recall status). https://www.nhtsa.gov/recalls

Corrections. Published 2026-07-22, corrected and fully re-measured 2026-08-12. Three changes worth flagging: (1) the article previously described Ford's completion rate as the lowest of any major automaker, which contradicted its own Table 3 — BMW, Nissan, Kia and Hyundai all sit lower, and Ford leads on pool size rather than rate; (2) the 149.5 million figure was described as unrepaired cars and is now correctly labelled unrepaired recall obligations, with the 450.3 million denominator published alongside it; (3) all completion figures were refreshed from a new snapshot, which moved several materially, most notably Ford's F-150 trailer-brake recall from 16.12% to 65.65%.

All figures: a Zilocar analysis of NHTSA "Recalls Data" (Socrata dataset 6axg-epim), N=10,494 completion-tracked vehicle recalls, data as of August 12, 2026.