Fact-checking began as a newsroom function and became an industry.
Hundreds of organisations now publish on claims circulating online.
Evaluations show that a correction does change beliefs, on average and modestly.
The old finding that corrections has not survived .
Two harder problems remain.
The first is arithmetic: false claims are cheap to produce and expensive to check.
A single checked article can take a day; a generated one takes a second.
The second is : corrections reach a fraction of the original audience.
Attention has therefore shifted from checking claims to labelling sources and slowing spread.
works surprisingly well: asking a user to open a link before sharing it reduces sharing of false stories.
, which teaches the technique rather than the claim, holds up better over time than a correction.
People who can recognise a false dilemma do not need to be told about each new instance.
Institutional trust, however, sets a on all of it.
A correction from a distrusted source is read as confirmation of the original claim.
That is why local, specific and boring corrections outperform national, general and dramatic ones.