Preprint on text-level Misinformation on Social Media

In a new preprint, Sami Nenno together with Kamil Fuławka and Philipp Lorenz-Spreen from our team and Cornelius Puschmann from the University of Bremen use a new text-based method to detect misinformation on social media. The study analyzes more than half a million posts from all German members of parliament and their official party accounts across Facebook, X, Instagram, and TikTok. Using a novel text-matching approach, the researchers link posts to over 5,000 fact-checking articles and 1,500 community notes. Compared to other methods, they identify ten times as many posts that contain misinformation.

Their findings show that misinformation remains relatively rare overall but is not evenly distributed. Instead, it clusters around certain parties and topics — in some contexts reaching misinformation rates of about ten percent. By going beyond link-based detection, this approach broadens the scope of misinformation research, providing a more detailed view of where and by whom misinformation is spread, and highlighting the central role of certain parties in its circulation.

Find the full preprint here: https://osf.io/preprints/socarxiv/p6yh9_v1/