Authorship Attribution of AI‑Generated vs. Human Written News Articles Using Machine Learning

In this study, news articles from journalists were compared to AI-generated articles that were designed to mimic human-written text. The AI-generated articles were identified using similar parameters and perspectives listed in the associated prompts. Specifically, the writings of four individual journalists were compared to articles generated by Large Language Models (LLMs) using multiple machine learning classification algorithms. The findings reveal consistently high levels of accuracy in distinguishing text written by human authors from text generated by AI. In addition, the machine learning models demonstrated the ability to correctly attribute authorship across both human-authored and LLM-generated articles. While additional improvements in classification accuracy will be necessary before written documents can be attributed to specific LLM platforms with certainty, the results of the current study provide promising evidence for future research. More specifically, the high levels of predictive accuracy observed in this study may serve as a foundation for further experimentation involving feature engineering, classifier optimization, and hyperparameter tuning. It is hoped that such efforts will further improve overall machine learning model performance and authorship attribution capabilities.