Good Car Bad Car is a small publication that covers a lot of ground: every brand and model in two countries every month, plus China, Europe and a global explorer, plus the news that explains the numbers. We can do that because software and AI tools do much of the collecting, checking and drafting, and people decide what is published. This page says exactly where the line sits, so you can judge our work with the facts in hand.
The numbers
Every figure on this site comes from a source we can name: automakers’ own sales releases, industry associations such as the CPCA and CAAM in China, and government registration data in markets where manufacturers do not report directly. Figures are reconciled against manufacturer totals, and we run standing audits that compare our tables with the sources and log every discrepancy until it is resolved. When a manufacturer revises its numbers, we revise ours.
The sales tables and the model, brand, automaker and ranking pages are generated from our database by software we built, and the short summaries on those pages are generated from the same numbers. They refresh as new data arrives. No number on this site is estimated by an AI model or typed in from memory.
How articles are written
Articles start with the data and with primary sources: the release, the filing, the transcript, the regulation. AI tools help us gather those sources, cross-check figures against them, draft the article and edit it against GCBC’s own editorial standards, which cover sourcing, balance, numbers and plain language. A person then reads the piece, checks that the numbers match the sources, chooses the illustration and decides whether it publishes. Articles do not publish themselves.
The tools we currently use include Anthropic’s Claude for research and drafting, with other models for specific tasks such as image generation. We name them because it is a fair question; we will change them when something better comes along, and this page will say so.
Articles are published under the Sariah byline, which is the name of GCBC’s research desk rather than a person; everything under it is produced the way this page describes.
What AI is never used for
- Inventing a number. If a figure is not in a source, it does not go in the article.
- Inventing a quote or a source. Quotations come from documents and statements we can link to.
- Describing an experience nobody had. We do not publish driving impressions or reviews; our articles are about what sold and why.
- Presenting a guess as a fact. Where something is an estimate, an outlook or our own interpretation, the wording says so.
The research desk
Sariah is the name GCBC gives its AI research desk: the scheduled systems that read manufacturer releases, regulatory filings and GCBC’s own database, draft and check pieces to the editor’s rules, and prepare them for review. Sariah is not a person. Every piece carries the desk’s name and the editor’s, because both did the work: the desk drafted and checked, the editor read, edited and decided. We say so here rather than hide it.
Images
Article images are stock photographs, manufacturers’ press images, and AI-created illustrations made to a written house style. Charts are drawn from the same data as the tables.
Independence
Nothing you read here was paid for by anyone. Advertising is sold and served by Raptive, and advertisers have no say in what we publish. Members pay for a better experience of the same data, not for different conclusions. Manufacturers do not see articles before publication, and we do not accept guest posts, sponsored articles or paid links. We have no political angle: the numbers are the numbers, whoever they favor.
Who is responsible
Good Car Bad Car is published by Leetone Enterprises LLC, a Colorado company. Eli Lemberg is the editor and publisher: he sets the editorial policy, decides what the site covers and how, reads every piece before it is published, and can hold, correct or withdraw any piece. He is accountable for everything on the site, whichever tools helped make it. If we get something wrong, tell us at [email protected] with the page and the figure in question. We check it against the source, fix it and say that we did.
What we collect about you as a reader is a different subject, and it is covered in our Privacy Notice. Terms for citing and licensing our data are on the About page.
Corrections are made in public and listed on the corrections page. How the figures are sourced, imported and checked is set out in more detail on the methodology page.
This page describes how the site works as of October 2026. When the process changes, this page changes with it.