Method
How we verify
AI accelerates collection and drafting. Every claim needs a source. A story publishes only when every gate passes.
- 1Collect — scan many sources continuously
- 2Cluster — group one event, remove duplicates
- 3Score — rate importance and source credibility
- 4Verify — check each claim against sources
- 5Gate — every check must pass, and high-risk topics wait for a human editor
- 6Publish — facts go live with confidence and an update log
Who publishes
The automated newsroom can publish a story without a human editor when every gate passes. The gates are: each confirmed claim is checked against its source text, at least two independent origins support the story, no high-risk topic is flagged for human review, and rights to the sources are cleared. Several outlets that repeat one wire report count as one origin.
High-risk topics always wait for a human editor. They are named individuals, minors, graphic content, market-moving price or trading claims, elections, active investigations, sanctions and war, defamation risk, medical claims, terror, self-harm, sexual crime, and stories that need a right of reply. A human editor reviews source rights, completes a standards and legal review, and records the right-of-reply status before such a story is published.
Editors can also approve and publish a story by hand.
Errors are corrected openly and logged on the article. Serious cases are retracted with a reader notice. This is the same for every story, however it was published. To report an error or ask for a right of reply, see the Corrections page or write to corrections@newsmesis.example.
We mark uncertainty clearly (Confidence: High / Medium / Developing), separate confirmed facts from unconfirmed claims, and correct visibly with an update log.