This research presents a pipelined approach to distinguish between human-written and AI-generated text using natural language features and classification models.

• Leverages prompt-based rewriting features inspired by RAIDAR methodology
• Incorporates content-based features from the NELA toolkit
• Evaluates effectiveness through comprehensive experiments on the Defactify4.0 dataset

This work addresses critical security concerns by providing tools to identify potential misinformation, fraud, or deceptive content created by AI systems. As LLMs become more sophisticated, robust detection methods become increasingly essential for maintaining information integrity.

SKDU at De-Factify 4.0: Natural Language Features for AI-Generated Text-Detection