How we research

A source-read comparison is not a product test. We make that distinction before the recommendation.

Start with the exact source

We use official documentation, manufacturer specifications and current pricing pages for the claims they can support. A regional model page or plan matrix matters more than a generic brand promise. We record source URLs and read dates on each article. A source may be corrected or superseded after that date.

Separate three kinds of statement

• Published fact: a labeled limit or feature attributed to its source.
• Editorial inference: why that feature may fit a stated job, not a measured superiority claim.
• Illustrative example: arithmetic or a suggested evaluation procedure, not results from testing.

Make uncertainty visible

If a feature matrix loses check marks, source pages disagree or a model is region-specific, we narrow the claim or say what must be confirmed. We do not invent missing dimensions, prices, scores or trial results. A date is not a promise that every account or region has identical terms.

Unranked shortlists

An alternatives shortlist needs three distinct sourced options. Named pairwise comparisons cover the named two products; no spurious third recommendation is needed. We do not award a universal best badge, sell positions or publish star ratings. The current collection is AI-assisted research and writing, not hands-on testing or human expert certification.

Money and corrections

See the link policy and corrections.