An Arabic Dataset of Human and AI-Generated Texts for Automatic AI Text Detection

The Arabic Human and AI-Rewritten Text (AHAIT) Dataset is a balanced Arabic corpus developed to support research on AI-generated text detection and Arabic natural language processing (NLP). The dataset contains 3,180 Arabic text samples, including 1,590 human-written texts and 1,590 AI-rewritten texts. Human-written texts were collected from Arabic news articles and academic documents covering multiple domains, including politics, sports, arts, international affairs, and academic writing. AI-rewritten texts were generated using ChatGPT and Gemini through a controlled rewriting process that preserves the original semantic content while producing linguistically natural Arabic text. The dataset includes text content, word-count statistics, source model information, and class labels. AHAIT can be used for AI-generated text detection, text classification, authorship analysis, stylometric analysis, and Arabic NLP research.