ARTIFICIAL INTELLIGENCE USE IN PROFESSIONAL BACHELOR THESIS WRITING: A SURVEY-BASED ANALYTICAL FRAMEWORK

Abstract

This paper examines students’ use of artificial intelligence (AI) in professional bachelor thesis writing and asks how those patterns can be described in a form useful for future support and governance in higher education. The study compared the 2024 and 2025 graduating classes at the same Lithuanian institution. The target population comprised 399 students, and the final sample included 258 respondents, with 129 from each graduating class. The questionnaire covered thesis-related tasks, named tools, perceived benefits and risks, attitudinal statements, and open-ended comments. Quantitative data were analysed in SPSS using frequencies, percentages, reliability analysis, Shapiro-Wilk tests, non-parametric comparisons, and Spearman correlations. Open-ended comments were grouped thematically. The results show a stable pattern: students used AI most often for paraphrasing and summarising, translation, grammar and style revision, and literature search. ChatGPT and Google Scholar clearly dominated the tool profile. Students valued AI mainly because it saved time and made writing easier to manage, but they also pointed to errors, weaker originality, and the need to verify AI-generated content. Lecturer guidance was the strongest attitudinal result, suggesting a practical need rather than a general wish for more regulation. The paper translates the survey evidence into a layered framework that separates student context, thesis tasks, named tools, perceived risks, and guidance expectations. This framework may inform later supervision dashboards, disclosure forms, or rule-based support systems for thesis administration. The study also identifies a limit in the available evidence: when reliable Lithuanian-language AI-detection data are not collected, self-reports and AI-use declarations should be treated as supervision and verification indicators, not as proof of misconduct.