The last panel at the ECREA 2026 conference in Brno starts with Morgan Wack, whose focus is on X’s Community Notes programme. Community Notes enables users of X to add notes to misleading posts, and these are then rated by other users; if these ratings show agreement even across partisan lines, the notes are publicly attached to posts, and reduce reposts and likes on the platform by nearly 50%.
YouTube, Meta, and TikTok are now implementing similar crowdsourced annotation approaches; this is part of their move away from professional and platform-funded fact-checking and moderation frameworks. But only 14% of all proposed notes ever appear publicly on X; it takes more than 26 hours for them to appear; and more than 40% of genuinely helpful notes are suppressed by the need to reach bipartisan agreement – and of course the process depends on the continued voluntary involvement of a broad range of users of the platform.
But does it actually matter which claims are fact-checked? The removal of easily debunkable claims can give them more support, yet these are also more likely to be fact-checked, since plausible claims are harder to debunk; so, this project explored how crowdsourcing participants rated the claims they fact-checked.
Claims that are easy to fact-check turned out to be more likely to appear with notes; claims that were harder were less likely to receive a Community Note. Harder claims do still receive notes, but received considerably less engagement. And worryingly, hard checkability and high plausibility tend to go together, so some of the potentially most problematic claims will go unchecked – only 4.6% of hard but plausible claims receive a Community Note.
One explanation for this is that effort scales with difficulty, while rewards do not. One solution might be to escalate stalled claims to professional fact-checkers; alternatively, to pay more for ratings of hard claims, or to make ratings visible; or finally, to nudge experience ragers when claims stall.
