UDC 37
DOI: 10.36871/ek.up.p.r.2025.05.07.022
Authors
Laura K. Khadzhieva,
Ahmed K. Chadaev,
Grozny State Petroleum Technical University named
after Academician M. D. Millionshchikov
Abstract
The aim of this study is to systematize and analyze existing methods for recognizing the probability of application of artificial intelligence (AI) systems in text documents. The objectives included studying linguistic, stylometric, statistical, and machine learning-based approaches to AI-generated content detection; identifying their advantages, disadvantages, and areas of application; and identifying key challenges and prospects for the field. The results demonstrate that despite significant progress, there is no universal and absolutely reliable detection method; the effectiveness of each approach depends on the complexity of the AI model, the amount of training data, and the specifics of the text being analyzed. The scientific novelty consists in a comprehensive review and comparative analysis of heterogeneous techniques, and the practical significance consists in the possibility of using the presented methods to combat misinformation, academic dishonesty and unfair use of AI technologies in the textual environment.
Keywords
artificial intelligence, text generation, text recognition, stylometry, machine learning, large language models, AI detectors, text authentication, content analysis
