Study of registered US voters in 2016: 6.7% of all news consumed on Twitter was fake news, 1% of users were exposed to 80.0% of it, and 0.1% shared 79.8% of it
Context & Ripple Effects
This study lands in the middle of an accumulating evidence base on how misinformation actually moves. Earlier work had already established that inaccurate news spreads faster and further than true stories and that humans, not bots, are primarily to blame, and that Facebook engagement with known fake-news sites fell by more than half after the 2016 election while Twitter shares kept rising. What this paper adds is the distributional answer: fake news was 6.7% of what registered voters consumed on Twitter, but exposure and sharing were extraordinarily concentrated — 1% of users saw 80% of it, and 0.1% of users shared nearly all of it.
First-order effects
- Twitter's moderation problem is now quantifiably a super-spreader problem: interventions aimed at a tiny fraction of hyperactive accounts address roughly four-fifths of fake-news sharing, making targeted account-level action far cheaper than platform-wide content policing.
- Follow-up research found the same pattern held in 2020, where just 2,107 voters accounted for spreading 80% of election fake news — meaning platforms could have identified and constrained these cohorts between cycles.
Second-order effects
- The Facebook–Twitter divergence sharpens: as Facebook's post-2016 measures cut fake-news engagement by half or more, Twitter became the residual high-sharing channel, putting its ad business and policy choices under disproportionate scrutiny from advertisers and regulators.
- Because [[a:934164|over 80% of the accounts that spread false information in 2016 remained active and were still pushing more than a million tweets a day]], enforcement that removes content but not repeat-offender accounts leaves the same network intact for the next election.
Third-order effects
- If the pattern holds across elections, platform governance shifts structurally from content-level moderation to actor-level intervention — identifying and constraining the small super-sharer cohort — which raises harder questions about account suspension criteria and political neutrality than fact-checking individual posts does.
- Concentration this extreme also reframes the 'fake news' problem itself: since most voters were barely exposed, the systemic risk lies less in mass persuasion than in a persistent, identifiable minority amplifying low-quality content within an ecosystem observers already describe as unhealthy.
The trend: Misinformation research is converging on a durable finding across election cycles — that fake news on social platforms is driven by a tiny, persistent super-sharer cohort — pushing platform governance toward account-level enforcement over content-level moderation.