LLMs across the Software Development Life Cycle in Project-Based Software Engineering Courses: an Empirical Study

Abstract

Large Language Models (LLMs) are increasingly used in software development and are becoming part of software engineering education. However, there is limited empirical evidence on how students use AI tools across development tasks in project-based Software Engineering courses. This paper presents an empirical study based on approximately 8,500 validated student-submitted AI usage logs collected over five semesters between 2022 and 2026 from two undergraduate courses at the National Technical University of Athens. We examine in which phases of the Software Development Life Cycle (SDLC) students use AI tools, how they evaluate AI assistance, and how their perceptions evolve over time. The results show that AI usage is mainly concentrated in coding and testing, with lower usage in requirements, architecture, and design. Students rate quality higher than knowledge gained, and those with higher task experience report higher quality and knowledge gains. Overall, students view AI positively, report low perceived threat, and tend to use it more for task completion than for learning.