What we publish
Raven Wire cross-checks other newsrooms’ reporting and official data. It does not fact-check stories itself or certify them. Software groups reports about the same event into one story, using the headline, place and time. Every story page lists all of its reports with links, the outlet, its tier and the time it was published.
How a story is scored
Each story gets a confidence score out of 100. The full breakdown is on the Method page:
- Independent confirmation (up to 45): separate newsrooms. Wire copy re-run by ten outlets counts once; outlets run by the same state count once.
- Source track record (up to 25): the best outlet’s tier and editorial independence.
- Wire service (15): an established newsroom runs AP, Reuters or AFP copy.
- Geographic spread (up to 15): outlets from different countries.
- Hedged language (minus up to 20): “reportedly”, “unverified”, “could not be confirmed”.
Verdicts
Verdicts are defined terms. They describe how many independent newsrooms we found, using the rules on this page. They are not a guarantee that a report is true or false.
- Confirmed: an official primary source (an agency, government body or operator’s own feed) backs it, and at least 2 independent newsrooms from 2 or more countries report it. This is the same two-key rule that decides alerts.
- Corroborated: at least 2 independent newsrooms report it, but no official source backs it yet. However many newsrooms carry it, it is not called confirmed without that second key.
- Developing: reports are still coming in.
- Single source: one newsroom so far.
- State media only: only state-controlled outlets carry it. A record of what a government says, never confirmation (score capped at 25).
- Disputed: a fact-checker rated a matching claim false or misleading (score capped at 20).
Search engines are only asked to index stories reported by at least two independent newsrooms.
Verification and alert levels
Big events also get an alert level from L1 (feed) to L5 (critical). Levels come from rules, not from AI: for measured events, from the source’s own fields (earthquake magnitude and USGS PAGER alert, GDACS colour, hurricane class, status-page impact); for stories, from an escalation word list, the coverage and the measured events they are linked to.
Verification decides what a level may do. We use a two-key rule:
- Confirmed: an official primary source and 2+ independent newsrooms from 2+ countries.
- Corroborated: 2+ independent newsrooms, no official source yet.
- Official: an official source only.
- Developing: one newsroom, or only state media. Shown at most as L3.
- Disputed: a fact-checker or our AI analyst doubts it. Shown at most as L2.
Only confirmed L4 and L5 events are eligible for alerts.
Sources and state media
Sources are tiered: T1 wire services and official bodies, T2 major newsrooms, T3 specialist and regional outlets. State-controlled outlets (for example RT, TASS, CGTN, Xinhua) are labelled STATE wherever they appear and never count toward confirmation. The full list, with each source’s tier, ownership and health, is on the Method page.
Fact-checks
When a published fact-check (ClaimReview) from a fact-checking organisation matches a story, we show it on the story page and apply the caps above.
AI
An AI analyst writes the plain-language read on top stories and the daily “Connecting the dots” brief. It sees only the reports in the story, it never sets a score or raises a level, and its text is always labelled. Details: AI use.
Bylines
Stories carry no personal bylines: they are assembled by software from other newsrooms’ reports, following the rules on this page. Each report is credited to its outlet on the story page.
Mistakes
If we group, score or label something wrongly, we fix it. See Corrections.