Data for Propaganda by Prompt: Tracing Hidden Linguistic Strategies in Large Language Models (Barfar & Sommerfeldt, 2025)

This tabular dataset was generated as part of the research titled "Propaganda by Prompt: Tracing Hidden Linguistic Strategies in Large Language Models (Barfar & Sommerfeldt, 2025)." It includes the data and metadata for over 340,000 AI-generated news articles produced using OpenAI’s GPT-3.5-Turbo, GPT-4o, and GPT-4.1 models, along with more than 119,000 corresponding human-authored data. The dataset is stored in a SQLite database (.db file) and includes the ai_prop_liwc table. To access the dataset, download the compressed (.zip) file, extract the .db file, and use the accompanying analysis tools provided in the linked GitHub repository: https://github.com/LeeSomm/PropagandaByPrompt