The standard is usefulness, evidence, and accountable release.

We do not publish to fill a daily quota. A topic must serve a defined reader and produce something useful that is not already covered elsewhere in the network.

Ideas that duplicate the same reader need, thesis, and artifact are rejected or assigned to one clear canonical article.

Factual claims should use current primary sources when they are available. Firsthand observations must be labeled as such. Unknowns remain unknown instead of being filled with convincing language.

We do not publish unsupported “best,” “leading,” guaranteed-result, guaranteed-booking, revenue, or performance claims. Recollective service descriptions are identified as first-party descriptions.

AI agents may organize research, outline, draft, edit, and check candidate articles. They may not invent sources, interviews, people, permissions, client results, or commercial facts.

Before publication, a Recollective-designated AI publishing supervisor reviews the exact article and destination against this policy. A passing supervisor decision may authorize publication without separate human review. This is a Recollective-controlled automated review, not independent or human editorial judgment. Recollective remains the accountable publisher and is responsible for the final text.

Drafted, source-checked, supervisor-approved for publication, published, and live-verified are separate states. Only the exact content approved by the designated AI publishing supervisor may be added to the publication allowlist. A successful draft, agent run, source change, build, or deployment is not publication proof.