On the Vulnerability of Community Notes to Participation Concentration

As platforms scale down professional fact-checking, community-based alternatives are promoted as more transparent and democratic. The leading example is Community Notes on X, where contributors write annotations and collectively rate their helpfulness under a consensus-oriented algorithm. Yet such systems are structurally prone to skewed participation, where a small minority of highly active raters dominates contribution dynamics. This raises a fundamental question: does the Community Notes algorithm provide meaningful robustness against such concentration? We address this question through a largescale empirical analysis of the full public dataset of notes and ratings (2021-2025). We characterize participation dynamics, showing that contribution activity is highly concentrated and that high-activity raters exhibit distinctive behavioral patterns: selective engagement, partisan leaning, and temporal activity patterns difficult to reconcile with careful manual evaluation. To assess the robustness implications, we directly run the public open-source production implementation of the Community Notes algorithm, enabling counterfactual simulations that measure how note visibility changes as hyperactive raters are selectively removed. Our results reveal that the system is structurally unstable: note visibility often depends on the behavior of a few dozen highly active raters, and minor perturbations in their participation produce markedly different outcomes, casting doubt on the integrity of its moderation decisions. Rather than decentralizing fact-checking authority, Community Notes concentrates effective control in a small, skewed subset of raters whose rating intervals are difficult to reconcile with careful manual evaluation, raising fundamental concerns about the output integrity of communitybased moderation systems.