DETECTING FACTUAL ERRORS AND FABRICATED SCIENTIFIC REFERENCES IN GENERATIVE AI RESPONSES: THEORETICAL FOUNDATIONS AND RESEARCH PROSPECTS AMONG UNIVERSITY STUDENTS IN UZBEKISTAN

Generative artificial intelligence (AI) tools are increasingly used by university students to obtain explanations, summarise literature, prepare assignments and search for scientific sources. However, these systems can produce, in a fluent and convincing style, incorrect facts, inaccurate bibliographic details and references to publications that do not exist. Drawing on literature on AI hallucinations, information literacy, evidence-based practice and cognitive psychology, this article analyses the theoretical foundations of detecting factual errors and fabricated references in generative AI output. Empirical studies show that a substantial proportion of references generated by large language models are fabricated or inaccurate, and that even medical trainees do not reliably recognise such errors. The article proposes a conceptual framework for studying students' ability to distinguish accurate from inaccurate information and genuine from fabricated references, a multicentre study design for universities in Uzbekistan, and a six-step educational model for improving AI literacy.