Purpose. This study investigates whether evaluation using artificial intelligence (AI) with a customized GPT model can be comparable to, or complementary to, human evaluation in graphic design projects created by master's degree students. Methodology. A total of 180 logos were evaluated over six years using a 10-item Likert scale rubric. Each project received a double assessment, one by human experts and one by AI. Results and Conclusions. The results show agreement in five categories between human evaluators and AI (e.g., originality, relevance) while, in another five categories (e.g., balance, design principles), AI tended to overvalue the works when compared to human assessment. The qualitative analysis of the AI’s feedback identified strengths, but also limitations. In conclusion, a hybrid human-AI evaluation could improve fairness and efficiency in creativity grading, although AI ...