Evidence first. Analysis labeled. Corrections visible.

• Primary records from developers, publishers, platforms, regulators, filings, and official channels are preferred.
• Qualified primary records are enriched with relevant approved reporting and official social records before autonomous publication is considered.
• A source link does not automatically make every claim verified; first-party confirmation, attributed reporting, context, and reaction are stored as different evidence classes.
• Rumors, estimates, and analysis are labeled and are not presented as confirmed facts.
• Anonymous or community claims require additional corroboration before publication.

Dossier scores use a documented lifecycle-aware method, bounded evidence influence, confidence reporting, and dated snapshots. Editorial judgment remains part of the method; the score is an assessment, not an objective promise of quality or commercial performance.

Automated systems may collect metadata and bounded excerpts from an approved registry of primary, reporting, and official social sources. AI may synthesize a claim-linked research bundle into a working report. Owner- authorized automation can approve only reports that pass the multi-source evidence, attribution, originality, official-media, duplication, and publication-limit gates. All other work remains private for manual review.

Advertising is labeled and separated from editorial controls. Sponsors and ad providers cannot purchase a score, suppress a correction, or approve coverage. Material commercial relationships are disclosed near the relevant content.