An IFCN fact-checker: multiplying verification capacity 6x
Illustrative scenario: a fact-checking team triages 300 claims a week and reserves reviewers for the claims that need original reporting.
This page describes a design target. No organisation has been named, quoted, or contracted.
The challenge
A fact-checking organisation with 14 researchers receives several hundred claims a week during election periods. Triage is manual: someone reads the queue, guesses which claims are checkable and consequential, and assigns the rest to a growing backlog.
The team is careful, which is the point. It would rather publish 40 well-evidenced checks than 200 rushed ones. But the bottleneck means important claims are checked late, and duplicate research is common across partner organisations.
The solution
The proposed workflow ranks claims by check-worthiness — consequence, reach, and verifiability — and retrieves candidate evidence with passages rather than bare links. Researchers see the sources that support and contradict a claim, plus the sources that were rejected and why.
Verdicts follow the IFCN-aligned scale. FALSE and MISLEADING findings require a named reviewer. The evidence chain exports as JSON or PDF, so a partner organisation can audit the work without repeating it.
The results
Targets, not results. The pilot would measure triage time, reviewer agreement, and the rate of overturned verdicts.
What’s next
A time-boxed pilot on one election cycle, with an independent review of 50 verdicts sampled before publication.
“Capacity is not the same as output. We want to spend reviewer time on the claims that deserve it.”
Not a real customer quote