Do community notes and crowdsourced fact-checking work?
Partially, and the evidence points to a consistent pattern: they are reasonably accurate but too slow, and their weakest point is exactly where misinformation does most of its damage.
How the mechanism works. Contributors write notes providing context on posts, and other contributors rate them as helpful or not. The distinguishing feature is the bridging algorithm: a note is only shown publicly if it is rated helpful by people who have historically disagreed with each other on previous notes.
Why that design is clever. It makes brigading much harder. A coordinated group cannot simply upvote a note into visibility, because the system requires approval across rating patterns, not merely a majority. It is an attempt to surface notes that are broadly persuasive rather than factionally popular — and it is a genuinely interesting piece of mechanism design.
What the research has found:
Accuracy is generally good. Studies assessing displayed notes against expert judgement have found them largely accurate and well sourced.
They reduce resharing. Posts with a visible note are reshared less, and authors delete them more often.
They shift belief in the claim among people who see them.
But speed is the fatal weakness. Research has consistently found that notes take hours to days to appear, while the large majority of a viral post's views occur in the first few hours. By the time context arrives, most of the audience has already seen the post without it.
Coverage is low. Most posts flagged never receive a displayed note, because the bridging requirement means many proposed notes never reach consensus. Notes on politically contested claims are least likely to be displayed — precisely the category where disagreement is highest.
Why platforms favour it anyway. It is cheap relative to professional fact-checking, scales with volume, is more transparent, and — significantly — moves the judgement away from the platform itself, which reduces accusations of institutional bias.
The realistic assessment: a useful supplement with a real effect on the posts it reaches, not a replacement for other moderation, and structurally unable to intervene during the window that matters most.