From viral tweets claiming garlic water could cure COVID-19 to users promoting dog parasite drugs as treatments for the virus and myths that vaccines make people “Bluetooth connectable,” misinformation spread like wildfire on Twitter (now X) during the pandemic. At the same time, many users took to the platform to debunk false claims and counter fake health advice.
But who were the users behind the misinformation spread and these competing streams of information?
In a new study, USC researchers developed an artificial intelligence (AI) tool to identify tweets containing COVID-19 misinformation and distinguish the users spreading false claims from those actively countering them.
Rather than simply detecting misinformation, the researchers aimed to understand the people behind it by classifying user profiles into two groups: accounts that spread misinformation and accounts that countered it.
Using a large dataset of tweets posted between 2020 and 2021, the team trained large language models (LLMs) to identify misinformation-related content and analyze how the two groups differed in their online behavior and profile characteristics. Led by USC Viterbi and USC Stevens professor Emilio Ferrara and his PhD student Eun Cheol Choi, the project concluded this week and resulted in the paper, “Angry but Accurate: Detecting and Profiling the Counter-Misinformation Ecosystem on Twitter,” which was accepted to ACM Hypertext 2026.
Ferrara directs the USC HUMANS Lab and holds joint appointments in the USC Viterbi School of Engineering and the USC Mark and Mary Stevens School of Computing and AI‘s Thomas Lord Department of Computer Science, as well as the USC Annenberg School for Communication and Journalism.
Users Fighting Fake News Are Angrier, While Misinformation Spreaders Have Fewer Followers #
The study found that users who counter misinformation tend to be more established on the platform, with older accounts and larger follower bases than those spreading misinformation.
The results also showed that counter-misinformation posts were often “angrier” and expressed stronger negative emotions than posts spreading misinformation–challenging the common assumption that misinformation itself is the more emotionally charged content.
Specifically, researchers found that users fighting misinformation exhibited higher levels of anger, sadness and disgust in their posts.
In contrast, posts supporting misinformation tended to be longer and expressed more surprise, while corrective posts were generally shorter and more direct.
The team noted that findings from the study also have implications for how social media platforms moderate content.
The researchers noted that content moderation systems on platforms such as Twitter may automatically flag or downrank posts that contain high levels of negative emotion. As a result, these systems could risk inadvertently suppressing legitimate posts fighting false claims that play an important role in the counter-misinformation ecosystem.
COVID-19 and the Rise of Fake News Online Inspired the USC Study #
The team described the COVID-19 pandemic as a “wild time” for public health, noting that the fear, uncertainty and desperation surrounding the virus fueled tens of thousands of unique false claims that circulated worldwide, creating a rich environment for researchers to study misinformation.
But the sheer volume of misleading posts was not the only factor that motivated the research.
“Our lab has been investigating misinformation spread on social media,” Choi said. “We were specifically interested in the counter-misinformation ecosystem during 2020 because the dangerous nature of COVID-19-related misinformation made it a critical period for understanding how people respond to and combat false information.”
The team emphasized that on fast-moving social media platforms such as Twitter, false health advice, medical myths and conspiracy theories can spread rapidly and put lives at risk.
Published on July 29th, 2026
Last updated on July 29th, 2026