Gaming Consensus: Coordinated Manipulation in Crowdsourced Fact-Checking

Crowdsourced fact-checking systems have been adopted by major social media companies such as X, Meta, TikTok and Google with the aim of combating misleading information at scale without relying on centralized editorial control. These systems have been developed around a common underlying concept: a bridging mechanism that identifies notes flagging misleading information when they receive support from people with different perspectives rather than simple majority support. To our knowledge the only publicly disclosed bridging algorithms deployed for fact-checking are based on matrix factorization, as deployed by both X and Meta, augmented with additional components addressing abuse, targeted manipulation, and contributor brigades. This work examines the core matrix factorization portion of these systems, presenting theoretical and empirical evaluations of the degree to which coordinated users could vote strategically by leveraging the latent representations to fabricate the appearance of synthetic consensus within the bridging mechanism. Using historic production data, we find that up to 10.7% of lower quality notes could be manipulated above consensus thresholds using less than 10 ratings. We complement these findings with a theoretical analysis, revealing counterintuitively that rating a note as ``Not Helpful'' can increase its helpfulness score, as well as a cost model quantifying manipulation effort. We have developed and deployed mitigations within X's Community Notes algorithm to address synthetic consensus.

Social media platforms like X, Meta and TikTok work to address misleading information without relying on central editorial decisions using a concept known as Community Notes, which lets ordinary users write short notes that add context or corrections to posts. Rather than showing notes based on simple majority votes, the systems only show notes widely when people who usually hold different opinions agree the notes are helpful. In this paper, we examined the method these platforms use to identify such broad agreement. We tested whether coordinated users could strategically vote to get the system to approve low-quality notes. Using historical data, we discovered that up to 10.7% of weaker notes could be pushed over the approval threshold with less than 10 ratings. We also found that marking a note as “Not Helpful” can in unusual cases raise its overall score. We further analyzed how much effort this kind of manipulation would require. To address these risks, we designed and have implemented stronger protections in X’s Community Notes system, making the crowdsourced fact-checking process more resistant to coordinated interference.