Human vs. Machine Deception: Distinguishing AI-Generated and Human-Written Fake News Using Ensemble Learning

arXiv cs.CL / 4/14/2026

💬 OpinionSignals & Early TrendsIdeas & Deep AnalysisModels & Research

共有:

Key Points

The paper studies how AI-generated fake news differs from human-written misinformation, focusing on linguistic, structural, and emotional signals.
It builds document-level features using sentence structure, lexical diversity, punctuation patterns, readability metrics, and emotion-related measures (e.g., fear, anger, trust, anticipation).
Multiple classifiers (logistic regression, random forest, SVM, XGBoost, and a neural network) are compared, and evaluation uses accuracy and ROC-AUC.
Results indicate that readability-based features are the most informative predictors, and AI-generated text tends to show more uniform stylistic patterns.
An ensemble approach that aggregates model predictions yields modest but consistent performance gains over individual models, suggesting robust distinguishability via text properties.

Abstract

The rapid adoption of large language models has introduced a new class of AI-generated fake news that coexists with traditional human-written misinformation, raising important questions about how these two forms of deceptive content differ and how reliably they can be distinguished. This study examines linguistic, structural, and emotional differences between human-written and AI-generated fake news and evaluates machine learning and ensemble-based methods for distinguishing these content types. A document-level feature representation is constructed using sentence structure, lexical diversity, punctuation patterns, readability indices, and emotion-based features capturing affective dimensions such as fear, anger, joy, sadness, trust, and anticipation. Multiple classification models, including logistic regression, random forest, support vector machines, extreme gradient boosting, and a neural network, are applied alongside an ensemble framework that aggregates predictions across models. Model performance is assessed using accuracy and area under the receiver operating characteristic curve. The results show strong and consistent classification performance, with readability-based features emerging as the most informative predictors and AI-generated text exhibiting more uniform stylistic patterns. Ensemble learning provides modest but consistent improvements over individual models. These findings indicate that stylistic and structural properties of text provide a robust basis for distinguishing AI-generated misinformation from human-written fake news.