How we test
Every conclusion should tell you what it rests on, where it applies and what remains unknown.
Separate research from testing
Official documentation and independent sources support product facts. Hands-on or API testing is identified only when a published record documents the tasks, software or model version, access channel and date. Missing evidence stays missing.
Publish the conditions
Model reports state the dataset version, configuration, sampling and scoring rules. Runs retain failures, timeouts, refusals and retries. Results apply to that test set and date; they do not establish a universal ranking.
Explain the cost
Plans include billing period, limits, seats, tax notes and verification date where known. Annual commitments stay visible. Recorded costs and estimates are kept separate, and missing amounts are never treated as zero.
Keep commercial links separate
Affiliate links carry a disclosure. Program approval and payment conditions are distinct from editorial judgment. Clicks are not evidence of a purchase or commission.
Keep revisions traceable
Product verification, testing and article revision dates describe different events. Updated drafts are reviewed before replacing the public version. Errors can be reported through the contact details on our About page.