When Machines Lie Differently: Detecting AI vs Human Fake News
Year: 2026
Type: conference-paper
Abstract: The rapid advancement of large language models has introduced a new class of AI-generated fake news that coexists with traditional human-written misinformation, raising fundamental questions about the detectability of different sources of deceptive content. In this study, we investigate whether the detectability of fake news depends on its origin by constructing two controlled binary classification tasks: distinguishing real news from human-written fake news and from AI-generated fake news. Using a feature representation that captures lexical diversity, readability, and emotional characteristics, and applying logistic regression, random forests, support vector machines, gradient boosting, neural networks, and an ensemble model, we evaluate performance using the area under the receiver operating characteristic curve (AUC). Across all models, we observe a consistent and substantial performance gap, with near-perfect AUC for AI-generated fake news and lower accuracy for human-written fake news. Feature-level analysis reveals that AI-generated fake news exhibits more uniform patterns in readability and emotional expression, reducing its overlap with real news. These findings highlight a critical asymmetry in misinformation detection and underscore the importance of accounting for the source of misinformation, and for the distinction between authorship and veracity, when designing detection systems.
Institution: Kennesaw State University
Language: English
Subfield: Sociology and Political Science
Field: Social Sciences
Domain: Social Sciences
Open Access status: hybrid
OpenAlex ID: W7171436353