Datasets of artificial intelligence (AI) model responses to tax assessment questions at the university second-year level

The rapid rise of artificial intelligence (AI) in higher education raises questions about its role in learning and assessment, particularly in taxation, where complex calculations, problem-solving, and professional judgment are required. This study analysed how accurately AI models could respond to second-year taxation assessments at the University of Pretoria using the Revised Bloom’s Taxonomy (RBT). A mixed-method approach was applied, combining quantitative accuracy checks against memoranda with qualitative RBT analysis of cognitive processes. The results showed that AI performed well on lower-order thinking tasks but struggled with higher-order skills, with instances of irrelevant or incorrect outputs (“hallucinations”). The findings highlight AI’s potential as a supportive tool in taxation education but also its limitations in complex reasoning, underscoring the need for careful integration alongside human expertise. The names appearing on the data files are fictitious.