A new study is offering a closer look at how disinformation differs from credible reporting—not just in accuracy, but in how stories are constructed and presented.
The research, Examining Narrative Patterns in Disinformation and Trustworthy News: A Comparative Analysis, analyzed 610 English-language articles, including both pro-Kremlin disinformation and verified news coverage spanning international events from 2015 to 2023. Using a combination of large language models and knowledge graph analysis, researchers evaluated how each type of content was structured across several narrative dimensions.
The findings show consistent differences in how disinformation is framed. Articles identified as disinformation were more likely to rely on conspiratorial narratives and show hostility toward media institutions, while also lacking diverse sourcing and strong evidence. They also scored lower in areas such as cultural context and long-term narrative consistency.
In contrast, trustworthy reporting demonstrated stronger use of evidence, broader sourcing, and more balanced narrative construction. One notable finding: emotional intensity did not significantly differ between disinformation and legitimate news, suggesting that both can use similar emotional cues even when the underlying narratives diverge.
Researchers also developed a “NarrativeRisk” scoring model to help distinguish between the two. The model showed a strong ability to differentiate disinformation from credible reporting, reinforcing the idea that narrative structure itself can serve as a useful indicator in identifying misleading content.
The study further identified three recurring narrative profiles—one dominated by disinformation, one by trustworthy reporting, and a third mixed category—highlighting the complexity of today’s information environment.
For those working in homeland security, counterterrorism, and information integrity, the findings underscore a growing challenge: identifying not just what information is false, but how it is being framed to influence audiences.
