Show the claim. Show the receipt.

Before a claim can publish, its subjects must resolve to canonical entities and its material facts must connect to stored evidence. A source label alone is not proof. Sensitive or conflicting claims require human review and cannot publish automatically.

• Confirmed: supported by authoritative evidence appropriate to the claim.
• Reported: attributed to a named independent source and not represented as direct confirmation.
• Rumor: unresolved information that cannot change canonical relationship status.
• Recap: an account of observable episode material with spoiler state attached.
• Opinion: commentary, interpretation, or a fan prompt—not a factual claim.

Material claims link to evidence records that preserve source identity, retrieval context, and the exact support used. Multiple reports about one event are clustered instead of published as duplicate stories.

A person image requires both subject-identity evidence and a permitted use basis. When either is unresolved, the correct fallback is a named non-person graphic—not another person who looks plausible.

Automation may organize evidence, draft text, or suggest metadata. Source pages are inert input, generated claims must resolve back to stored evidence, and no automation may impersonate the owner or silently replace submitted owner text.

Substantive corrections create a visible version and audit record. Retractions leave a clear correction shell while removing the item from feeds, sitemaps, advertising, and current-content surfaces.

Read the correction process ↗