CivitasPoint
Editorial standards
CivitasPoint helps readers examine competing claims, challenge assumptions, and investigate evidence themselves. We have no predetermined political conclusion. We apply consistent scrutiny to competing viewpoints without giving unsupported claims equal credibility.
Four labels, clear expectations
- News: primarily reports verified, attributed developments.
- Analysis: explains evidence, context, tradeoffs, and competing interpretations.
- Opinion: presents judgments, recommendations, or arguments.
- Explainer: provides background and explains how something works.
Sources before certainty
Link material factual claims to primary documents, official data, or credible reporting. Identify dates, units, geographic scope, and revisions. A source’s assertion is not automatically an established fact. Corroborate disputed claims and describe unresolved uncertainty.
Three distinct activities
- AI-assisted factual review: retrieving sources and checking whether they support specific claims, dates, quotations, figures, and context.
- AI-assisted cross-review: another AI tool challenges reasoning, framing, omissions, and potential errors.
- Publisher approval: the publisher decides whether an article is ready to publish.
AI agreement is not independent confirmation. Different tools can repeat the same mistake or rely on the same underlying source. Deep analysis alone does not establish that a claim has been verified.
Our process does not include an independent human fact-checking team. AI review can miss errors, and readers are encouraged to report corrections.
Publishing standard
Factual claims must be checked against accessible, relevant sources. When a consequential claim cannot be verified, we qualify it, omit it, or hold the article for further review. We do not describe a claim as verified merely because multiple AI tools agree.
Questions must be grounded in evidence. We do not use a question mark to introduce an unsupported accusation. Headlines, images, and social posts must not imply more than the evidence establishes.
Fair argument
Represent the strongest relevant objection. Distinguish correlation from causation, an announcement from implementation, and a projection from an observed outcome. Apply the same scrutiny across political parties, including arguments we initially find appealing. Do not manufacture balance or uncertainty.
Accountability and corrections
The publisher makes publication decisions. Publication approval does not, by itself, establish that a factual claim has been verified. This standard is not a claim that every review step has been completed for every legacy article.
Correct material errors transparently with a dated note stating what changed. Label sponsored material and disclose relevant conflicts. Do not invent staff, qualifications, review history, interviews, or firsthand observations. Do not use AI-generated images as documentary evidence of real events.
Corrections policy · AI policy · Contact information
Update — October 5, 2026: Positioning and workflow disclosures were revised to describe the single-publisher, AI-assisted process and its limitations.