How do fact-checkers actually work?

Fact-checking is a research process, and understanding the method makes it easier to judge whether a given check is any good.

Two distinct activities share the name:

Internal fact-checking happens before publication. Some magazines employ checkers who verify every factual assertion in a piece against primary sources before it runs. This is rarer than it was and rarer in newspapers than magazines.

External fact-checking is what most people mean: organisations examining public claims after they are made, and publishing verdicts. Full Fact, Snopes, PolitiFact and AFP Fact Check are examples, and many are signatories to the International Fact-Checking Network's code, which requires transparency of sources, funding and methodology.

The method, broadly:

Isolate a checkable claim. "The government spent £X on Y" is checkable. "This policy is a disaster" is not — value judgements and predictions cannot be fact-checked, and good checkers say so rather than rating them.

Trace it to a primary source — the original dataset, the full statement in context, the study itself rather than the press release about it.

Consult independent experts where interpretation is required.

Publish the working, so a reader can follow the reasoning and disagree with it. A verdict without sources is an assertion, not a check.

Its genuine limitations:

Rating scales compress nuance. Many claims are partly true in ways "Mostly False" does not capture, and the choice of rating involves judgement.

Selection matters. Which claims get checked is itself an editorial decision, and a recurring criticism is uneven attention across the political spectrum.

Corrections travel poorly. Research consistently finds the original claim reaches far more people than the correction.

Some evidence of a backfire risk exists, though the effect appears weaker than once believed.