Toker, SacipAkgun, Mahi2025-07-062025-07-0620251470-32971470-330010.1080/14703297.2025.25142422-s2.0-105007609939https://doi.org/10.1080/14703297.2025.2514242https://hdl.handle.net/20.500.14411/10664This study examines whether assessments focused on higher-order cognitive skills can help reduce AI-driven plagiarism in educational settings. A total of 123 participants completed three tasks of increasing complexity, aligned with Bloom's taxonomy, across four groups: control, e-textbook, Google, and ChatGPT. Results from repeated-measures ANOVA revealed that both similarity scores and AI plagiarism percentages significantly declined as task complexity increased (p < .01). The ChatGPT group initially exhibited the highest AI plagiarism rates during lower-order tasks, but their performance improved on higher-order tasks requiring analysis, evaluation, and creation. These findings highlight a clear distinction between similarity scores and AI plagiarism detection, emphasising the need for combined evaluation methods. Overall, the study demonstrates that designing assessments to foster higher-order thinking offers an effective strategy for minimising plagiarism associated with generative AI tools, providing practical implications for academic integrity policies and instructional design.eninfo:eu-repo/semantics/closedAccessAi PlagiarismBloom'S TaxonomyChatGPTGenerative AiTask ComplexityArticleQ3Q1WOS:001504595400001