Generative AI and the Quality of Student Research Projects

The article presents a comprehensive analysis of the impact of generative artificial intelligence systems on the quality of student research work. Drawing on empirical data from the 2025 summer examination period, including observations and expert evaluations of course and graduation papers, as well as findings from the Higher Education Policy Institute survey, the authors describe real practices of AI use by students and faculty. Typical usage scenarios are identified and compared with existing research on the properties of large language models. The philosophical and theoretical section develops a three-level analytical framework (epistemic, instrumental, and normative), demonstrating which procedures of knowledge justification remain essential and which cognitive and interpretive operations must be retained by humans. The study highlights key risks: the illusion of understanding and stylistic coherence replacing epistemic soundness, a growing crisis of source reliability, a decline in critical thinking skills with prolonged tool dependence, unequal access to technology, and ambiguity in assigning responsibility for research outcomes. At the same time, the authors note several positive effects of AI, including enhanced literature review capabilities, support for structuring arguments, and assistance in text editing, provided that adequate procedural safeguards are in place. As a result, the paper proposes a set of practical quality assessment criteria (such as substantive originality, procedural transparency, reproducibility of reasoning, and explicit declaration of AI contribution), a mandatory reporting system (including a declaration of AI use, a working portfolio with several intermediate drafts and a list of prompts, and an explanatory note), regulated verification and sanction procedures, and dedicated educational modules for students and instructors. The authors conclude that a combination of transparent rules, training programs, and institutional mechanisms enables the integration of generative tools in ways that strengthen rather than substitute genuine research practices.