AI-generated text detection: A comprehensive review of methods, datasets, and applications

Abstract

This review examines the rapidly evolving field of AI-generated text detection, which has gained critical importance following the widespread deployment of advanced large language models like ChatGPT. We analyze the technical foundations, methodological approaches, evaluation frameworks, and practical applications of detection technologies designed to distinguish between human and machine-authored content. The paper synthesizes current knowledge across key dimensions: detection techniques ranging from statistical approaches to neural architectures, datasets and their limitations, performance metrics and evaluation challenges, real-world implementations across educational, publishing, and legal domains, and emerging research directions. Our analysis reveals significant challenges, including the inherent adversarial nature of detection, cross-domain generalization difficulties, and fairness concerns regarding certain writer populations. We identify promising trends toward multi-scale analysis, human-AI collaborative frameworks, and complementary provenance-based approaches. The review concludes that effective detection remains feasible but requires combining multiple approaches, domain-specific customization, and attention to ethical implications. This comprehensive examination serves as a resource for researchers, practitioners, and policymakers navigating the complex technical and societal dimensions of AI text detection in an era of increasingly sophisticated generative AI systems.