Scientific Text Detection Tools: Accuracy, Bias, and Academic Integrity Challenges in Multilingual Educational Environments

Abstract

The widespread use of artificial intelligence (AI) has significantly influenced academic writing and raised increasing concerns about research integrity. As AI writing tools have become more common, universities and academic institutions have adopted AI text detection systems to distinguish human-written text from AI-generated content. However, questions remain regarding the accuracy, fairness, and reliability of these tools, particularly when evaluating texts written by non-native English speakers. This study aims to examine the accuracy, limitations, and potential bias of current AI text detection systems through an analytical literature review. Peer-reviewed articles, preprints, and institutional reports published between 2023 and 2025 were systematically reviewed. The analysis focused on detection accuracy, false-positive and false-negative rates, linguistic bias, and the main limitations of existing detection methods. The findings indicate that current AI detection tools still suffer from inconsistent accuracy and are prone to errors, particularly when assessing writing produced by non-native English speakers. The review also highlights the vulnerability of many detection systems to paraphrasing and other text modification techniques. It is concluded that AI detection tools should not be used as the sole basis for academic judgment. Instead, their use should be supported by human evaluation, clear institutional guidelines, and continuous improvement to ensure greater fairness, accuracy, and reliability.