What Fact-Checking Actually Catches

The errors that verification work catches are rarely dramatic. Nobody expects to find a fabricated interview. What actually turns up, more often than not, is smaller and stranger: a job title that changed eighteen months ago and was never updated, a date pulled from a press release that quietly contradicted the court record, a spelling of a name that two separate documents disagree on.

None of these are the kind of error that would sink a story on its own. But add up four or five of them across a single piece, and a pattern starts to matter — not because the reporting was careless, but because drafts move fast and small facts get carried forward from note to note without anyone circling back to check them against the original source.

The claims worth checking first

Not every sentence in a draft needs the same scrutiny. We tend to prioritise anything that could be independently confirmed against a document — a date, a figure, a title, a location — over claims that rest on a single source's characterization of events, which need a different kind of handling entirely. Numbers, in particular, travel badly. A figure quoted from a report often shifts slightly by the time it reaches a third or fourth draft.

• Names and titles, checked against the most recent official record
• Dates, cross-referenced across at least two independent sources
• Figures and statistics, traced back to their original document
• Direct quotes, confirmed against notes or recordings where available

What we return, and what we don't resolve

A verification pass ends with a written sheet, not a verdict. For each claim, it notes what was checked, what confirmed it, and — just as usefully — what couldn't be resolved in the time available. That last category matters. Being honest about an unresolved claim gives the reporter a real choice: dig further, soften the language, or cut it, rather than publishing on an assumption that quietly turned out to be wrong.