Car Recall Scraper — NHTSA Recalls, Complaints & VIN

Scrape official US vehicle safety data from NHTSA — the federal regulator — in four modes: recalls, owner complaints, crash-test star ratings and VIN decoding.

This is the vehicle half of product safety that the CPSC does not cover: cars, trucks, tyres and child seats are regulated by NHTSA, not the Consumer Product Safety Commission.

No API key required.

Four datasets, one actor

Mode	What you get
Recalls	Campaign number, component, hazard consequence, remedy, manufacturer, and NHTSA's park it / park outside fire warnings
Owner complaints	What owners actually reported: crash and fire flags, injury and death counts, incident dates, component, partial VIN
Crash-test ratings	Overall, front, side and pole star ratings, rollover risk, plus recall, complaint and investigation counts
VIN decode	40+ factory specs from a single VIN: engine, transmission, plant, airbags, driver-assist systems, battery capacity

What you get — recalls

Field	Description
campaignNumber	NHTSA campaign ID, e.g. 23V430000
reportReceivedDate	When the recall was filed, normalised to ISO
component	The failing system, e.g. SERVICE BRAKES, HYDRAULIC
summary	What is wrong and which vehicles are affected
consequence	What can happen as a result
remedy	What the manufacturer will do about it
parkItWarning, parkOutsideWarning	NHTSA's most severe warnings — do not drive, or do not park indoors
overTheAirUpdate	Whether the fix ships as software

What you get — complaints

Field	Description
odiNumber	Complaint ID
dateComplaintFiled, dateOfIncident	Both normalised to ISO
crash, fire	Whether the incident involved either
numberOfInjuries, numberOfDeaths	Harm reported
component, summary	System affected, and the owner's own words
partialVin	Partial VIN as submitted

Example input

{
  "dataType": "recalls",
  "make": "honda",
  "model": "accord",
  "modelYear": "2022",
  "maxRecords": 100
}

Find only serious complaints:

{
  "dataType": "complaints",
  "make": "ford",
  "model": "f-150",
  "modelYear": "2023",
  "injuriesOnly": true
}

Decode VINs into full specs:

{ "dataType": "vin", "vins": ["1HGCV1F34LA012345"] }

Example output

{
  "recordType": "recall",
  "campaignNumber": "23V430000",
  "reportReceivedDate": "2023-06-15",
  "manufacturer": "Honda (American Honda Motor Co.)",
  "component": "SERVICE BRAKES, HYDRAULIC:ANTILOCK/TRACTION CONTROL:CONTROL UNIT/MODULE",
  "summary": "A ball valve in the vehicle stability assist (VSA) modulator may leak brake fluid...",
  "parkItWarning": false,
  "overTheAirUpdate": false
}

Who uses this

• Dealers and used-car marketplaces — check every VIN in your inventory against open recalls before you list it
• Product-liability and injury lawyers — complaint volume by component is how a defect pattern is established, and the death and injury counts are right there
• Insurers and actuaries — crash-test ratings and complaint density by model year
• Fleet and logistics operators — catch a park it recall on a vehicle your drivers are using today
• Automotive journalists and researchers — track which components generate the most owner complaints across a model's life
• Vehicle history and inspection products — VIN decode plus recalls in one call

A date-format trap this actor handles for you

NHTSA serves two different date orders from the same host, which is a genuine source of silent corruption if you read the API directly:

• /recalls returns ReportReceivedDate as DD/MM/YYYY — 15/06/2023 is 15 June
• /complaints returns its dates as MM/DD/YYYY — 09/28/2026 is 28 September

Parse both the same way and roughly a third of your dates land in the wrong month. This actor normalises both to ISO YYYY-MM-DD, and falls back to the other reading when a value is impossible under the expected one, so filedAfter works consistently across record types.

The API is also inconsistent about its results key — results on recalls and complaints, Results on VIN decode and safety ratings. Handled.

Notes on coverage

• Scope is the US market. A model sold only outside the US will not be found.
• Crash-test ratings exist only for tested body styles. NHTSA rates specific configurations, not every trim, so a lookup can legitimately return nothing for a model that does have recalls. The actor says so rather than failing.
• Recalls are per make/model/year, so a campaign covering several models appears once per model you query.
• Complaints are unverified owner reports, not findings. High complaint counts are a signal worth investigating, not proof of a defect — treat summary as a claim.
• VIN decode works on 11–17 character VINs; partial VINs return partial specs, and anything NHTSA cannot interpret comes back in errorText rather than being silently dropped.

Pricing

Pay per result. Each recall, complaint, rating or decoded VIN counts as one result, and the crash/injury and date filters are applied before charging.

Notes

• Data comes from NHTSA's public recall, complaint, SafetyRatings and vPIC APIs.
• This is an unofficial actor and is not affiliated with NHTSA or the US Department of Transportation.