guarledes

What is on the site

Model years 2010 to 2027, 2,053 of the vehicles carrying at least one federal defect investigation, and 448 manufacturers in the recall archive. Every figure on this page is counted from the data at build time, so none of them can drift away from what the site actually holds.

Where the data comes from

Four federal sources, all read from the files themselves rather than from anybody's copy of them. What each one is, and what it does not cover, is set out on data sources.

• Complaints — the ODI bulk file, every complaint since 1995, rebuilt from scratch each refresh (published October 2, 2026).
• Recalls — the campaign API, walked by campaign number rather than by vehicle, because the bulk recall file was withdrawn on 19 August 2026 and the per-vehicle endpoint cannot find a recall on a vehicle you have not already thought to ask about.
• Investigations — the ODI investigation file, which is what says whether the agency is still looking.
• Crash test ratings — NHTSA's own NCAP results, where a configuration was tested.

What this site will not publish

The rules below cost the site pages, and they are the reason to trust the pages that remain. Each is checkable against the site itself.

• A state whose law could not be read at the source. 50 of 51 jurisdictions have a lemon-law page here. Arkansas does not, because its statute is published through a system that will not serve an automated reader. Every fifty-state table that fills that gap is a summary copied from another summary, and they disagree with each other. Rather than repeat numbers nobody here has read at the source, the page names the Arkansas Attorney General, Public Protection Department and leaves the row out.
• A recall campaign NHTSA holds only as a number. A campaign gets its own page when the record carries the defect, the consequence and the remedy. Older and withdrawn campaigns that carry none of those stay on the vehicle pages that aggregate them; a heading over three empty sections is not a page.
• A vehicle with too little on file to say anything. Model years before 2010 are out, and below 30 complaints a vehicle needs a recall or an open investigation to earn a page, with an absolute floor of 8. The dataset holds far more vehicles than the site publishes.
• A number the data cannot carry. Where a join comes back empty the sentence is left out rather than softened. A complaint that names the same system as a recall is not evidence of the same defect, and the pages count complaints and say so instead of implying otherwise.

How mistakes are caught

A counting rule that is wrong is wrong on thousands of pages at once, and the failures that matter here are silent ones: nothing throws, every page renders, and the first report would be a Search Console notice weeks later. So the build refuses to publish rather than warning.

• Structured data is checked against the real schema.org vocabulary, after every recall page shipped a type that does not exist.
• Internal links are checked against the pages the build made, after the recall pages shipped 2,046 links to vehicles the publish rule had excluded.
• Every page in the sitemap is checked for a route to it from the homepage, after 2,088 recall pages turned out to be reachable from nowhere at all.
• A collapse in page or image count stops the publish, so a truncated download cannot replace a working site with a stub.

The counting rules themselves are written down in how this is counted rather than left for you to trust.

What it will not tell you

Whether your car is safe. A complaint is one owner's unverified account of what happened to them; a pattern across thousands of owners is not a diagnosis of the vehicle on your drive; and a model that sold in larger numbers collects more complaints without being worse. The only authoritative answer for one particular car is a VIN check against NHTSA's own records, which every page here links to and which this site does not try to replace.