AI-Generated Content Detection Model Using Zero-Shot Learning Algorithm

Abstract

The rapid evolution of generative Artificial Intelligence (AI) models has revolutionized digital content creation, yet it has also raised significant ethical and practical concerns regarding authenticity, misinformation, and academic integrity. Existing detection methods, which largely depend on supervised machine learning with labeled datasets, struggle to keep pace with the adaptability of emerging AI models. This study proposes a novel AI-generated content detection system leveraging Zero-Shot Learning (ZSL), which eliminates the reliance on task-specific labeled data while improving adaptability and scalability. The system integrates text preprocessing, linguistic feature extraction, and a ZSL-based classification framework that categorizes content as human-written or AI-generated, supplemented with a confidence scoring mechanism for transparency. Experimental evaluations against benchmark models, including RoBERTa, DistilBERT, and the GPT-2 Output Detector, demonstrate that the ZSL model achieves highly competitive results, with accuracy rates exceeding 94% for human-written text and 99% for AI-generated text. The findings highlight ZSL’s potential as a cost-effective, flexible, and future-ready approach for AI content detection across domains such as education, journalism, and social media. This work underscores the need for scalable detection frameworks that can adapt to the rapidly advancing landscape of generative AI.