Study: Humans Spread COVID-19 Misinformation More Effectively Than Bots on X

A new arXiv preprint from Lynnette Hui Xian Ng, Wenqi Zhou, and Kathleen M. Carley (Carnegie Mellon University) investigates how low-credibility narratives about COVID-19 vaccines spread structurally on social media. Analyzing 5.8 million messages from X across three temporal stages — Pre-Vaccine, Vaccine Launch, and Post-Launch — the team introduced two novel metrics: Appeal (network-weighted popularity of a message) and Scope (an author's message popularity-weighted network penetration). The study aimed to compare the influence of automated accounts (bots) versus human users in propagating misinformation.

The results reveal a striking pattern: in all timeframes, human-distributed low-credibility narratives achieved significantly higher structural influence than bot-generated ones. Moreover, a conditional temporal effect emerged — human-driven narratives peaked in both Appeal and Scope during the critical Vaccine Launch week, while automated accounts only maximized their impact during the highly uncertain Pre-Vaccine period. This suggests that organic human engagement is more potent at shaping belief networks during moments of high public attention, whereas bots exploit early uncertainty. The findings have implications for platform moderation strategies and vaccine communication campaigns.