Editorial standards

Sources and provenance

Every editorial item keeps its source list attached and public. We summarize and link; we do not republish other outlets' work. Items ingested from official feeds name the feed and the retrieval time.

Automation is labeled — always

Some briefs are drafted with AI assistance from linked source material. Those items carry a visible “AI-assisted summary” label plus the provider, model, and prompt version used. Automation is never presented as original reporting, and automated drafts about crime, elections, allegations, or minors can never publish without two human approvals.

Risk-based review

• Green — routine official notices and calendars. May auto-publish only from explicitly whitelisted official sources.
• Amber — regional reports, development, explainers, ordinary submissions. Requires human approval.
• Red — crime, elections, allegations, named misconduct, minors in sensitive contexts. Requires two independent approvals and conflict checks. Never automated.

Every editorial override records a reason code in the audit log.

Corrections and retractions

Material changes create a public changelog entry on the article and in the site-wide corrections log. Retracted articles keep their URL with a clear notice — we don't memory-hole mistakes.

Independence

Agencies and officials can contribute through labeled official accounts, but they hold no editorial or moderation authority. Criticism of officials, institutions, advertisers, editors, or moderators never affects a member's standing by itself. Editors and moderators declare conflicts and recuse; the transparency page publishes aggregate outcomes.

MH Impact relevance test

Regional stories run only with a material Mountain House connection. Each MH Impact item must answer: what happened, why it matters here, who is affected, and what happens next — with sources. Location alone is not relevance.