How coverage earns trust

• Every factual claim used in an article must map to a linked source.
• Official statements, independent reporting, and rumor-tier material are distinguished rather than blended into one confidence level.
• Rumors are attributed and visibly framed as rumors; they are never presented as established fact.
• At most two short quotes may be used from any one source, always with attribution and a link. Facts may be reported; another writer's expression may not be copied.

Drafts are written from a structured fact brief containing attributed claims, a timeline, entities, uncertainty, and open questions. The drafting step does not receive raw source prose beyond the limited excerpts and brief quotations allowed by policy. This boundary is part of the anti-plagiarism control.

• Citation coverage for every factual claim used.
• Source-fidelity checks for names, numbers, and material assertions.
• Similarity checks against source expression.
• Visible rumor framing whenever rumor-tier material contributes.
• Image provenance and credit from permitted, licensed sources only.
• A meaningful information-gain finding.
• The daily publication cap and all applicable risk holds.

Automatic publishing is a runtime setting that defaults off. Even if an owner later enables it, only configured eligible formats that pass every gate may publish automatically. Why-it-matters analysis, community pulse, controversy explainers, evergreen guides, and all risk-flagged material remain human-reviewed. The current posture and eligible format list belong on the AI disclosure page.

A named human editor must be published before activation and is responsible for editorial decisions, review outcomes, corrections, and this policy. The current identity status appears on the editor page.

Leaked, embargoed, or datamined information is summarized with a link at most and always held for human review. Tragedy-adjacent coverage, legal disputes, and health or financial claims are also human-only. Passing a format rule never overrides these holds.

Reddit RSS is used only as low-frequency lead radar, not bulk commercial sentiment mining. Discord monitoring is excluded. SteamDB may be used as a human reference and is not scraped. Community discussion can prompt a reporting question, but it does not become a verified fact without independent sourcing.

Ads are absent at launch. Any future ad, affiliate relationship, free product, or sponsorship will be labeled where readers encounter it. Commercial arrangements do not determine coverage conclusions, source confidence, or correction decisions.