Key facts
- 18.1% (1,018 of 5,619) of US car and light-truck recall campaigns filed 2010-2026 prescribe a software remedy rather than a part.
- 32.2% (148 of 459) of those filed in 2025 - the peak year, and roughly four times the 2012 rate of 8.2%.
- 10.7% is what the same classifier returns on NHTSA's unrestricted "vehicle" file. That basis is 59% RVs, buses, heavy trucks and trailers, which almost never carry a software fix, so the familiar "about one in ten" framing is a dilution artefact. This is the check that moved our own headline.
- Only 2.6% of car recall campaigns say the fix arrives over the air. Most software recalls still require a dealer visit.
- 16.2% of software-remedy campaigns also say "replace", so the software-versus-part split is a spectrum, not a binary.
- Ford 177 campaigns, the most of any filer; Tesla 36.6%, the highest within-maker share among the 20 makers with 40+ campaigns.
- No completion data exists. The Part 573 flat file has no remedy-status field, so neither we nor any VIN check can tell you a flash was installed.
Do software recalls still need to be done on a used car?
Yes. A software recall - including an over-the-air one - is a prescribed fix, not a self-healing one. It stays an open recall until the update is actually applied, and on a used car the earlier owner may never have had it done. The repair method says nothing about whether the repair happened.
"It's just a software recall" gets treated as if it fixes itself. It does not. Even an over-the-air update depends on the car receiving and accepting it, so owner action, connectivity and model-year capability all matter - and on our numbers only 2.6% of car recall campaigns offer over-the-air delivery in the first place. The other software recalls need the same dealer appointment a broken part would. A software recall can also sit open for years; the biggest campaigns tend to be the least likely to be fixed before resale, so a software label is no reason to relax.
The only reliable confirmation that a specific car's recall was performed comes from a franchised dealer, by VIN.
How often is a recall fixed with software instead of a part?
In 18.1% of US car and light-truck recall campaigns filed since 2010 - 1,018 of 5,619 - and in 32.2% of those filed in 2025.
That 2025 figure is the one worth holding onto: for cars filed last year, software was the prescribed fix in about a third of all recalls.
A note on a number you may have seen elsewhere, because it is the check that changed our own answer. Run the identical classifier on NHTSA's unrestricted recall file - every record flagged RCLTYPECD = V, which is what "vehicle recalls" usually means in a published figure - and the share drops to 10.7%. Nothing is wrong with the arithmetic; the basis is wrong for a car buyer. V means vehicle, not car. Of the 13,743 campaigns on that basis, 8,124 (59%) are filed by RV, bus, heavy-truck, trailer, motorcycle and specialty builders - Forest River (625 campaigns, 6 with a software remedy), Keystone RV (193, none), Thor Motor Coach (131, two). Those campaigns are real recalls, and they dilute a software-share denominator almost to half its honest value.
So the car-only rate is nearly double the all-vehicle rate, and both are in the frozen dataset.
Table 1 - software-remedy share of US car and light-truck recall campaigns, by filing year (every year shown, no selection):
| Filed | Campaigns | Software remedy | Share | Over-the-air |
|---|---|---|---|---|
| 2010 | 174 | 15 | 8.6% | 0.0% |
| 2011 | 144 | 9 | 6.2% | 0.0% |
| 2012 | 171 | 14 | 8.2% | 0.0% |
| 2013 | 200 | 22 | 11.0% | 0.0% |
| 2014 | 325 | 43 | 13.2% | 0.3% |
| 2015 | 323 | 50 | 15.5% | 0.0% |
| 2016 | 350 | 43 | 12.3% | 0.0% |
| 2017 | 345 | 46 | 13.3% | 0.0% |
| 2018 | 366 | 54 | 14.8% | 0.0% |
| 2019 | 401 | 51 | 12.7% | 0.2% |
| 2020 | 350 | 51 | 14.6% | 1.4% |
| 2021 | 433 | 74 | 17.1% | 2.1% |
| 2022 | 392 | 88 | 22.4% | 5.6% |
| 2023 | 410 | 94 | 22.9% | 3.9% |
| 2024 | 447 | 119 | 26.6% | 6.3% |
| 2025 | 459 | 148 | 32.2% | 7.2% |
| 2026* | 329 | 97 | 29.5% | 9.4% |
2026 is partial, through the newest filing date in the file (RCDATE 2026-11-03), so its share is not comparable with a full year.
The rise is large but it is not monotonic, and we are not going to call it that. The share fell in 2016 (15.5% to 12.3%) and again in 2019 (14.8% to 12.7%). Showing only 2012, 2015, 2020, 2023 and 2025 would produce a clean staircase and a false impression; the full series above is the honest picture. Over-the-air delivery, by contrast, is close to monotonic, from zero before 2014 to 9.4% of 2026 filings.
One distinction to avoid double-counting: this measures how a recall is fixed, not what it was about. Our earlier read on how much of the recall mix is software and camera defects classifies the defect component instead, and the two are genuinely different measures. Today's hook is the proof: 26V717000 is a brake recall by component, and a software recall by remedy.
Which automakers fix the most recalls with software instead of parts?
Two different "most" claims, so both are given. By count of software-remedy campaigns, Ford leads with 177, then Chrysler/FCA and General Motors. By share of a maker's own recalls, Tesla is highest at 36.6%.
Both are ranked by campaign, never by units, and that restraint matters: a single Tesla or Ford over-the-air campaign can cover a million vehicles, so a units-weighted table would say something quite different. We make no units-weighted claim.
Table 2 - car and light-truck makers by count of software-remedy campaigns (N = 5,619 campaigns, retrieved 2026-11-08):
| # | Manufacturer (as filed) | Software | Of which software-only | Broad set | Total campaigns | Share |
|---|---|---|---|---|---|---|
| 1 | Ford Motor Company | 177 | 151 | 187 | 858 | 20.6% |
| 2 | Chrysler (FCA US) | 134 | 104 | 139 | 596 | 22.5% |
| 3 | General Motors | 91 | 87 | 95 | 524 | 17.4% |
| 4 | Mercedes-Benz USA | 71 | 69 | 72 | 389 | 18.3% |
| 5 | Volkswagen Group of America | 54 | 40 | 54 | 349 | 15.5% |
| 6 | Hyundai Motor America | 45 | 32 | 47 | 238 | 18.9% |
| 7 | BMW of North America | 44 | 43 | 44 | 379 | 11.6% |
| 8 | Toyota Motor Corporation | 40 | 30 | 54 | 321 | 12.5% |
| 9 | Nissan North America | 37 | 32 | 37 | 261 | 14.2% |
| 10 | Jaguar Land Rover NA | 35 | 31 | 37 | 133 | 26.3% |
| 11 | Tesla | 34 | 31 | 35 | 93 | 36.6% |
| 12 | Kia America | 33 | 18 | 33 | 169 | 19.5% |
Table 3 - car and light-truck makers by SHARE of their own recalls fixed with software (the 20 makers with at least 40 campaigns; all 42 are in the CSV):
| # | Manufacturer | Share | Campaigns | Software-only share |
|---|---|---|---|---|
| 1 | Tesla | 36.6% | 34/93 | 33.3% |
| 2 | Volvo Car USA | 27.5% | 22/80 | 25.0% |
| 3 | Jaguar Land Rover NA | 26.3% | 35/133 | 23.3% |
| 4 | Chrysler (FCA US) | 22.5% | 134/596 | 17.4% |
| 5 | Ford Motor Company | 20.6% | 177/858 | 17.6% |
| 6 | Mazda Motor Corp. | 20.6% | 21/102 | 14.7% |
| 7 | Kia America | 19.5% | 33/169 | 10.7% |
| 8 | Porsche Cars North America | 19.3% | 23/119 | 16.8% |
| 9 | Hyundai Motor America | 18.9% | 45/238 | 13.4% |
| 10 | Mercedes-Benz USA | 18.3% | 71/389 | 17.7% |
Read the last column before quoting the ranking, because it is where "instead of parts" gets weakest. For Kia, 33 campaigns prescribe software but only 18 do so without also saying "replace" - the share nearly halves, from 19.5% to 10.7%. Chrysler falls from 22.5% to 17.4% and Toyota from 12.5% to 9.3%. For Tesla, BMW, General Motors and Mercedes-Benz the two columns are close, so for those makers "fixed with code rather than a component" is close to literally true.
Two grouping caveats. Manufacturer labels are taken exactly as filed, so "Chrysler (FCA US, LLC)" is Stellantis and no cross-brand parent rollup was done - Audi and Porsche are not folded into Volkswagen, and Daimler Trucks is excluded from this universe entirely while Mercedes-Benz USA is in it. Rolling parents up would move ranks. And the allow-list is a judgement: 42 named manufacturers, every one listed in the frozen CSV so the boundary can be argued with rather than guessed at.
What is GM campaign 26V717, and is my Cadillac Lyriq or Chevrolet Blazer EV in it?
GM campaign 26V717000, filed 29 October 2026, recalls 67,380 vehicles because the brake system control module "contains a software defect that may cause a loss of braking during certain driving conditions". NHTSA's stated consequence is blunt: "A loss of braking ability increases the risk of a crash." The remedy, verbatim: "The BSCM software will be updated through an over-the-air (OTA) update or by a dealer, free of charge." Owner letters are expected 14 December 2026, and VINs became searchable on nhtsa.gov on 29 October 2026.
Affected models, exactly as NHTSA's record carries them:
| Model | Model years |
|---|---|
| Cadillac LYRIQ | 2023-2027 |
| Cadillac CELESTIQ | 2025-2026 |
| Cadillac VISTIQ | 2026-2027 |
| Chevrolet Blazer EV | 2024-2027 |
⚠️ NHTSA's record carries no trim or drivetrain restriction for this campaign, so do not rule your car out by trim. Some coverage describes the recall as limited to a performance drivetrain option, and to particular Blazer EV trims. That narrowing does not appear anywhere in the federal recall record or the recall API, we could not verify it at a primary source, and it is the kind of detail that wrongly reassures an owner whose VIN is covered. The VIN lookup at nhtsa.gov/recalls is the authority. Note also that NHTSA did not classify 26V717000 as do-not-drive or park-outside.
On the "more than 80,000" figure some outlets used: 67,380 is the brake-software recall alone. A separately filed GM campaign covers roughly 12,700 2025-2026 Chevrolet and GMC trucks for a steering defect whose remedy replaces a steering gear assembly - a hardware repair. The two should not be added into one "software recall" number: one is fixed by code, the other by a part. Press accounts also cited a complaint and incident chronology from GM's Part 573 filing; we did not extract those counts ourselves, so treat them as reported rather than as our figures.
Why was the same Lyriq brake software recalled in both 2024 and 2026?
Because a software fix does not permanently close a risk - and this is the clearest evidence in the dataset for the whole article.
GM filed campaign 24V589000 on 8 August 2024: 21,469 2023-2024 Cadillac LYRIQs, the same brake-control-module component, with the remedy delivered "through an over-the-air (OTA) update or by a dealer". Owner letters went out 3 September 2024. Two years later, 26V717000 sweeps those same 2023-2024 LYRIQs back in, inside a population three times the size.
Same model, same component, same delivery method, two campaigns. An over-the-air flash in 2024 did not end the matter, which is a concrete reason to confirm that an open software recall was actually performed rather than assuming an earlier update settled it.
NHTSA's own machine-readable record is no help here either: the agency's overTheAirUpdate flag reads false for 26V717000 despite that explicit over-the-air remedy text. That is not a one-off - we measured it across the file and NHTSA's over-the-air recall flag has been switched off since May 2025, and re-checking on today's fresh pull it still has not resumed: of the 31 car and light-truck campaigns filed in 2026 whose remedy text says the fix is over the air, the flag is set on none.
What a VIN check can and can't tell you here
A VIN check - a paid history report or NHTSA's free tool - can show that a recall is present and count open recalls. It cannot show whether the recall was remedied, flashed or completed. That is a limit of the public data, not of any one provider: the Part 573 file we classified has no completion field at all, only the CORRECTIVE_ACTION text describing the fix that was prescribed. Even a "closed" campaign status does not prove your specific car was flashed, as our read on "recall closed" versus an open NHTSA investigation sets out.
| Question | A VIN check (paid or NHTSA) | Who actually confirms it |
|---|---|---|
| Is a recall present on this VIN? | Yes - presence and count | - |
| Is the prescribed fix software or a part? | Partly - from the campaign's own remedy text | NHTSA campaign record / Part 573 |
| Was the recall actually remedied or flashed? | No | Franchised dealer by VIN; NHTSA owner lookup after the dealer reports it |
| Is the remedy delivered over the air? | Not reliably - NHTSA's structured flag has been dark since May 2025 | The campaign's remedy text; the dealer |
| Is there an open NHTSA investigation (PE/EA)? | No | NHTSA ODI |
| Per-unit firmware or flash status | No | Dealer |
A note on the policy backdrop, since it moved this week: NHTSA is consulting on modernising recall notification and on over-the-air remedies, and has been offering states funding to lift recall-repair completion rates. Those are proposals and programmes, not enacted rules - over-the-air recall modernisation is a direction of travel, not today's law, and none of it changes what a VIN check can prove today.
How to check open recalls on a used car by VIN
- Start free and authoritative. Enter the VIN at nhtsa.gov/recalls, which returns open, unremedied recalls for that specific vehicle.
- Then call a franchised dealer with the VIN and ask whether each open recall was performed. The dealer's system is the only place a per-VIN remedy status is confirmed - and that applies to a software flash exactly as it does to a replaced part.
- Screen the rest of the car's history in the same pass. A broader VIN history check surfaces accident, odometer, salvage and listing history alongside recall presence.
A Zilocar VIN check is one such option: it screens for recall presence and count and surfaces a used car's accident and airbag-deployment records, odometer and rollback check, theft (NICB) records, junk and salvage auction records, ownership history, sales-listing history, specs, NHTSA and IIHS safety ratings, and market valuation. It does not confirm a recall was remedied, track an NHTSA investigation, show per-unit firmware status, or return a legal title brand - those stay with NHTSA, the dealer, and NMVTIS or the state DMV. Use the VIN check to learn the car's history and see which recalls exist; use the dealer to confirm each one was actually fixed.
Method
Source. NHTSA Office of Defects Investigation Part 573 recall flat file, FLAT_RCL_POST_2010.zip, from static.nhtsa.gov/odi/ffdd/rcl/. Retrieved 2026-11-08: 15,099,486 bytes, HTTP Last-Modified 2026-11-08 08:03:41 GMT, expanding to FLAT_RCL_POST_2010.txt at 312,829,632 bytes, member timestamp 2026-11-08 08:02, testzip clean. 29 tab-separated fields, 247,176 rows. Newest RCDATE in the file is 2026-11-03. The single overTheAirUpdate figure quoted above comes from NHTSA's public recall API (api.nhtsa.gov/recalls/campaignNumber), queried 2026-11-08.
Deduplication. The file carries one row per make/model/year, so rows are collapsed to one record per CAMPNO: 15,382 distinct campaigns, of which 13,743 carry RCLTYPECD = 'V'. Where a campaign's rows differ in remedy text, the longest text is kept.
Universe, and why it is not all of V. · V means vehicle, not car: it carries RVs, buses, heavy trucks, motorcycles, trailers, fire apparatus and industrial equipment. The universe is therefore restricted to campaigns filed by an explicit allow-list of 42 manufacturers that build passenger cars and light trucks, giving N = 5,619 campaigns. Every one of the 42 is named in the frozen CSV, so the boundary can be inspected and disputed. A name-pattern classifier was tried first and abandoned because it leaked in both directions - it dropped car filers and kept International Motors, Altec, Pierce fire apparatus and several RV brands. Two label traps worth naming: "Hyundai Translead" is a semi-trailer builder, not Hyundai's car arm, and Daimler Trucks North America is excluded while Mercedes-Benz USA is included.
Classification. A campaign is software-remedied when its CORRECTIVE_ACTION text matches software, reprogram/re-program, firmware, reflash/re-flash, over-the-air, or OTA as a standalone token. This STRICT set is the published basis.
The keyword check that changed the basis. A BROAD set additionally matching calibrate, recalibrate, flash, ECU and PCM returns 1,064 (18.9%). Reading all 46 campaigns it adds on the car basis (102 across the unrestricted file) showed they include remedies that physically replace a damaged ECU ("dealers will replace the ECU", "replace any damaged ECU") and tyre-pressure-sensor recalibrations - the opposite of the claim this article makes - so BROAD was rejected as the headline basis and is published beside it instead. The two differ by 0.8 points, which is the useful result: the finding does not depend on how loosely "software" is drawn.
The mixed-remedy column. 165 of the 1,018 STRICT campaigns (16.2%) also contain "replace", i.e. software and a part, or "update or replace as necessary". The CSV carries a software_only_no_replace count per maker, and it is reported in Table 3 because it is where the "instead of parts" framing weakens - most sharply for Kia (19.5% to 10.7%).
The basis check that moved the headline, reported either way. On the unrestricted 13,743-campaign V basis the same STRICT classifier returns 1,477 (10.7%) against 18.1% on cars and light trucks, because 8,124 of those campaigns (59%) come from RV, bus, heavy-truck, trailer and specialty filers that almost never prescribe software. The rankings survive this and the rate does not: Ford/Chrysler/GM hold the top three by count, and Tesla/Volvo/Jaguar Land Rover the top three by share, under STRICT, BROAD, software-only, and on both the restricted and unrestricted bases. Note the inverse of what we found on an earlier census, where a unit-weighted figure survived a scope restriction and a campaign count nearly halved: which of a filter's statistics survives is not predictable, so each has to be checked on the restricted basis.
Known limits. (1) Text matching defines the universe, so a software remedy described in other words is not counted and 18.1% is a lower bound. (2) The file has no completion or remedy-status field, so nothing here says whether any recall was performed, by anyone, on any car. (3) Campaign counts are not weighted by units, and a units-weighted view would differ substantially; POTAFF is in any case a manufacturer launch-time estimate, not a count of cars on the road. (4) 2026 is partial. (5) The allow-list is a judgement call, published in full for that reason. (6) Makers with fewer than 40 campaigns are excluded from the share table because the ratio is unstable there; they remain in the CSV, flagged.
Frozen dataset. All 42 manufacturers, one row each, with total campaigns, the strict, broad and software-only software-remedy counts, the over-the-air count, all three shares, and a flag for whether the maker qualifies for the share table: software-remedy share by manufacturer, 2026-11-08 (CSV)
