Study: verified accounts on Twitter shared more content from deceptive sites than ever in 2020, interactions with deceptive content on FB also at all-time high
Context & Ripple Effects
This study lands mid-arc in the misinformation measurement debate. After 2016, platform pressure appeared to work on Facebook: engagement for sites identified as publishing fake news fell by more than half between the election and mid-2018, even as shares on Twitter kept climbing (that earlier decline). But by October 2020, researchers counted likes, comments, and shares from routine misinfo outlets tripling since Q3 2016 to 1.8B interactions (the rebound), and this report extends the pattern to a new peak.
What is new here is the vector: verified Twitter accounts — the accounts users are told signal authenticity — shared more content from deceptive sites in 2020 than ever before, while Facebook interactions with deceptive content also hit an all-time high. Verification, the platform's own trust marker, was amplifying the problem rather than filtering it.
First-order effects
- Twitter's verification badge loses credibility as a quality signal: the accounts most trusted by users were among the top distributors of deceptive-site content in 2020, undermining the badge's core promise.
- Facebook's post-2016 moderation gains are shown to have reversed — deceptive-content interactions at an all-time high contradicts the trajectory its earlier engagement declines suggested.
Second-order effects
- Advertisers face a brand-safety problem on both platforms: paid placements sit alongside distribution networks where deceptive sites reach record audiences, giving brands leverage to demand stricter adjacency controls.
- Fact-checking organizations and news publishers compete against deceptive sites that now out-engage them — the later NYU finding that misinfo publishers drew six times Facebook's engagement versus sources like WHO shows trustworthy outlets losing the attention market outright.
Third-order effects
- If platform-native trust markers (verification badges, ranking systems) systematically amplify low-quality publishers, the burden shifts to external institutions — regulators, advertisers, and independent researchers — to audit engagement economics rather than take platform self-reporting at face value.
- The recurring pattern across 2017, 2018, 2020, and 2021 studies points toward engagement-based recommendation being structurally biased toward deceptive content regardless of policy cycles — meaning moderation interventions produce temporary dips, not durable fixes.
The trend: Platform misinformation is moving through boom-and-bust cycles — each moderation push produces a measured decline, followed by record highs as deceptive publishers adapt and trusted vectors like verified accounts carry the load.