Generative Artificial Intelligence (AIGC) technology offers new opportunities for innovation in design education, yet its efficacy and potential risks at the elementary education level require systematic validation. This study focuses on first-year design students and employs a mixed-methods research approach (experimental-control group comparison, quantitative analysis, and qualitative feedback) to explore the impact of AIGC tools on the mastery of color composition knowledge and learning behaviors. The experimental group (n = 40) utilized KIMI and Dreamina AI tools for color scheme generation and iterative optimization over a two-week course, while the control group (n = 40) followed traditional teaching methods. The results indicate that the experimental group achieved a significantly greater improvement in standardized test scores (ΔM = 12.85) compared to the control group (ΔM = 6.5, p < 0.001), demonstrating the positive effects of AIGC on knowledge construction (t = 4.21), contextual learning (t = 5.34), and self-reflection (t = 3.76). Qualitative analysis further reveals that AIGC activates a “generation-revision-internalization” learning cycle through dynamic generation and instant feedback mechanisms. However, challenges such as technical adaptability (41% of students reported operational complexity) and educational ethical risks (18% plagiarism rate in solutions) were identified. The study proposes a “dynamic scaffolding” model, recommending a gradient-based AI intervention strategy and a multimodal ethical monitoring mechanism to balance technological empowerment and competency development goals. These findings provide empirical evidence for the optimization of AIGC educational tools and teaching practices, advancing paradigm innovation in constructivist theory within intelligent educational contexts.