How We Rate

The principle: only sources we actually read

Every entry in our database — every weakness, every cost figure, every part number, every battery value — comes from a web page that was actually read during research. Whatever cannot be documented stays empty. There are no estimated costs, no "typical" mileages and no problems that merely sound plausible.

We work with AI tools, and we say so openly: they search, read and summarise sources under fixed rules that a human sets and checks. The rules are what matters: a research run in which no source was read is discarded, no matter how convincing the result looks.

Data sources, ranked by reliability

1. Authorities — the German KBA recall database, NHTSA (USA), ADAC breakdown statistics, the TÜV report, DEKRA. Recalls are only included with a reference number.
2. Make and model forums — motor-talk.de, skodacommunity.de, goingelectric.de, peugeottalk.de, tx-board.de and many more. This is where real cases with mileage, workshop invoice and part number live. Forums that require a login are read with our own account.
3. Manufacturers, workshop and trade sites — data sheets, press releases, the ADAC catalogue for technical data and warranties, motorzentrale.de and raw Wikipedia tables for engine codes, parts dealers with OE numbers for prices.
4. Trade press — auto motor und sport, AUTO BILD, autozeitung, ADAC road tests; for the driving-fun texts and as cost examples.
5. Aggregators such as carwiki or gebrauchtwagenberater are used only as a starting point, never as the sole evidence.

We do not use content farms, used or salvage prices as new-part prices, AI summaries in search engines, or blanket statements about a make that prove nothing for a specific model generation.

What counts as evidence

• Weakness: at least two independent sources, or several different affected owners in a large forum. A single case in a single thread is an isolated case and does not enter the database.
• Repair costs: an invoice, a quote, or part price plus labour from the source. There are no ranges without a source.
• Engine codes and power data: double-checked against engine-family tables and motorzentrale.de; for power, displacement and production years the German KBA type data is authoritative.
• Battery data: replacement costs from manufacturer price lists, parts catalogues or workshop invoices; cell chemistry only where it is explicitly stated for the generation; measured ageing only from fleet measurements with a sample size (AVILOO, Geotab, TÜV), never from manufacturer claims.
• Driving-fun rating: at least one road test read for the specific engine and gearbox variant.

Before publication we check for duplicates, separate series problems from isolated cases and verify that every source URL is reachable. An automatic validator rejects entries without costs, search terms or a source. Whatever fails is not softened — it stays out.

Rating Scale: Severity

Overall Engine Rating

The overall engine rating (Good Choice, Caution, or Stay Away) is automatically calculated from all documented weaknesses. Severity and probability of each weakness flow into the score.

We do not set ratings manually. An engine with many minor weaknesses can still be rated "Good Choice", while an engine with a single critical weakness may fall to "Stay Away".

Editorial Process

The buying guide at the top of each model page puts the researched weaknesses into context: what is really critical, which engine is the best choice, what to check on a test drive. These texts are produced with AI assistance strictly from the documented data of that generation; only engines, recalls and costs that exist in the database with a source are mentioned. Approval, the rules and the correction of errors rest with the operator (see imprint).

Found an error or a missing weakness? Let us know.