This paper outlines a methodological approach for detecting potential plagiarism among student submissions in HTML and JavaScript-based programming tasks during exams by using artificial intelligence. We used OpenAI's ChatGPT to perform the plagiarism analysis. Detailed instructions about what to focus on were provided to ChatGPT, together with the prepared data sets for evaulation. The model performed a syntactic and structural comparison of three distinct assignment sets, analyzing code similarity using normalization techniques and visualizing the results through similarity matrices. This approach avoids penalizing functional similarity, instead focusing on structural and syntactical duplication since similar functionalities are expected as all students are solving the same task. We discuss the results which proved effectivness of this AI-assisted metodological approach in indentifiyng structurally similar submissions. Finally, we also analyze limitations of this method, issues that were encountered and we offer qualitative observation of usefulness of this approach for identifining students' unethical behavior in educational settings.

Keywords: Plagiarism detection, code similarity, ChatGPT, programming assignments, higher education.