Evidence Standards Policy
Defines how claims are sourced, graded, preserved and challenged so search and AI systems can understand why an article should be trusted.
Last reviewed Aug 4, 2026
Defines how claims are sourced, graded, preserved and challenged so search and AI systems can understand why an article should be trusted.
Evidence hierarchy
Prefer direct testing, official records and documentation, standards, regulator/court material, peer-reviewed research and attributable first-party statements. Credible independent reporting can corroborate or contextualise. Forums, reviews and social posts may reveal experiences but rarely prove general performance.
Claim-source linkage
Material claims require a source or test record linked at claim level in the editorial system. Records include title, publisher, URL or file reference, author, date, access date, relevant version/jurisdiction and preserved excerpt or snapshot where lawful.
Risk calibration
Higher-risk claims—security, privacy, money, legal obligations, Web3, safety or irreversible actions—need stronger and preferably multiple sources, explicit limitations and expert/editor review. Absence of evidence is not evidence of absence.
Quotation and context
Quotes must be accurate, proportionate and not misleadingly excerpted. Paraphrases preserve meaning and attribution. Statistics include population, period, method and uncertainty where relevant. Marketing claims are labelled and independently checked where possible.
Evidence labels and freshness
The six review labels are controlled vocabulary. Sources have review/expiry triggers. Broken links do not automatically invalidate preserved evidence, but material claims must be rechecked when product, policy, ownership, version or jurisdiction changes.
Auditability
Editors preserve sufficient evidence for a reviewer to reproduce the reasoning without exposing confidential sources or dangerous details. Conflicts and unsupported claims block publication.
Evidence states
Every Nerd Mango article carries one or more evidence states in its Evidence Passport. They describe how the article's findings were actually established. More than one can apply. Source-based states never imply hands-on use or testing.
Primary sources verified
Key factual claims are checked against official, first-party sources — vendor documentation, regulators, standards bodies or the primary party's own statements. It does not mean anyone used or tested the product first-hand.
Multi-source synthesis
The explanation is synthesised and cross-checked across several independent sources rather than relying on any single source. It does not mean the products were compared through hands-on testing.
Reported — not officially confirmed
Some points are reported by credible outlets or third parties but have not been confirmed by the primary party. They are flagged as reported, not verified.
Hands-on for this revision
A named person used the current product, service or account while preparing this specific revision of the article. It does not mean a formal, repeatable benchmark was run.
Controlled test
Findings come from a defined, repeatable test with a stated method, environment and date. It does not mean every configuration or edge case was covered.
Long-term experience
Reflects sustained real-world use of the product or service over an extended period, not a one-off session.
Specialist review
A domain specialist reviewed the article for technical accuracy, safety and fair framing before publication. It does not, by itself, mean the product was tested.
Not tested
No hands-on testing or controlled benchmark was performed for this article. It is built from sources, documentation and analysis. It does not mean the information is unverified — it means the basis is documentary, not first-hand testing.