Frequently asked questions
Answers about where the data comes from, how often it updates, why sales are not included, how to cite the site and how to use the data programmatically.
Are the complaints verified?
No. NHTSA publishes owner reports as filed, with personal details removed. Neither NHTSA nor this site checks whether an individual report is accurate. Patterns across many reports are meaningful; single reports are anecdotes.
Why not adjust for how many cars were sold?
Reliable sales figures by model year are not public in a form that can be joined to NHTSA data. We adjust for vehicle age and compare within brand, model year and segment, but none of that corrects for fleet size: a model that sold ten times as many cars will collect more reports. Read the index as report volume, not as a defect rate, and compare within a segment rather than across the whole site.
The data is American. Is it relevant elsewhere?
Many nameplates on this site are also sold in other markets, but engines, equipment and suppliers often differ from country to country, so a problem area may or may not carry over. Use the data as a guide to what to look for, and check recalls with your own national authority.
How often is the site updated?
Once a month. Every page shows the date of the newest complaint in the data. The what's new page lists how many complaints and recalls were added since the previous edition.
What does "typical" or "far more" mean?
They are bands of the complaint index: far fewer below 0.50, fewer 0.50–0.79, typical 0.80–1.24, more 1.25–1.99, far more 2.00 and above. Details on the methodology page.
Why does an older model year have more complaints?
Because it has been on the road longer. That is why the main figure is complaints per year on the road, and why trends are measured on the first two years of each model year only.
Is any of this written by AI?
Partly. The statistics are computed by fixed code, and the analysis paragraphs on data pages are generated from templates: fixed sentences filled with computed figures, the same way a weather report is written. No AI model is used when the pages are built, so data pages carry no AI-generated summaries. The guides and in-depth articles, and their translations, were drafted using an AI model (Anthropic's Claude) and have not been reviewed page by page by a person; the data blocks inside the articles are recomputed by code every month. Apart from owner complaints and recall and investigation summaries quoted from NHTSA records, nothing on the site is copied from other sources.
Can I use the data?
Yes. The underlying records are U.S. government public data. Our computed figures are available at /api/index.json and described for AI systems at /llms.txt. Please cite the site and NHTSA.
I found an error.
Use the contact page and include the page URL and the ODI or campaign number.