How Verdict researches used cars

Claims are ranked by how directly the source supports them. In descending order:

1. Official safety records. Recalls, defect investigations and complaint volumes from the U.S. National Highway Traffic Safety Administration (NHTSA). Recalls are described by campaign and affected build range, never as "every car of this year".
2. Manufacturer and catalogue specifications. Engines, drivetrains, trim availability and production years, cross-checked against NHTSA's vehicle product information catalogue (vPIC), which is also how generation boundaries on this site were verified.
3. Widely reported owner experience. Failure modes that recur consistently across owner reports and service discussion. These are written as patterns — what tends to go, roughly when, and how to spot it — never as a percentage.

Where a claim circulates widely but does not survive a check against the first two categories, Verdict says so explicitly in an "overstated concerns" section rather than quietly repeating it.

Every problem in a Verdict guide is written to the same shape:

• What it is — the component or system, in plain language.
• Which cars it affects — the specific model years, engine, drivetrain or market. A fault on one engine is not presented as a fault of the model.
• Symptoms — what the driver notices.
• What to check — what a buyer or mechanic can look at before committing.
• How serious it is — whether it is a walk-away, a negotiating point, or routine ageing.

Recall campaigns are kept separate from ownership wear, and manufacturer-acknowledged defects are kept separate from general maintenance items, because they mean very different things to a buyer.

Model years are grouped by generation, because most meaningful differences arrive with a platform, engine or transmission change rather than with a calendar year. Within a generation, individual years are called out only where something genuinely changed — a fix applied mid-cycle, a new engine, a recall confined to particular build dates.

"Best years" and "years to avoid" verdicts follow from that evidence, and are stated with their reasoning attached so you can disagree with the reasoning rather than having to trust a ranking. Where the evidence does not support a clear winner, Verdict says there isn't one instead of manufacturing a verdict.

Buying reports are generated by a large language model at the moment you request them, using the vehicle details you provide plus the listing content when you paste a link. That is what makes an instant free report possible, and it is also the report's main limitation: the model reasons from general knowledge about the vehicle, not from an inspection or from a records database. Figures in a report are estimates.

Research guides are drafted with AI assistance and then structured, fact-checked against the source hierarchy above, and edited before publication. Nothing is published simply because a model produced it.

Each guide stores its own last-updated date, changed only when the content changes, so the date means something. Guides are revised when new recalls or investigations are published, when a generation ends, or when a reader reports an error. Corrections are welcome via the contact page.

• Verdict has not seen the car. Condition, history and maintenance are unknown to it.
• Price and cost figures are estimates that vary by region, condition and season.
• Recall applicability depends on the VIN, not on the model year alone.
• Coverage is US-market focused; other markets got different engines, equipment and recall campaigns.
• No guide replaces an independent pre-purchase inspection.