Sensitivity is essential in art and Design education, traditionally cultivated through observation. With advancements in AI technology, new educational approaches are emerging. AI can visualize abstract concepts expressed in language; however, its ability to accurately represent intangible elements such as emotions and impressions remains unclear. This research investigates the impact of AI-generated abstract expressions on human sensitivity, focusing on Eroticism, a key aesthetic element in art and Design. Eroticism, deeply connected to human instincts, evokes strong emotions through color, form, and texture, making it significant for educational applications. We conducted a survey to identify words associated with Eroticism and selected 15 prompts. Using DALL·E 3, we generated images and collected impression evaluations from 113 participants. The consistency between prompts and images, as well as impression variation, was analyzed. The results classified the images into four categories based on impression variation and consistency. Concrete prompts (e.g., “naked,” “body”) showed high consistency and low variation, while abstract prompts (e.g., “mystety,” “softness”) had lower consistency. However, terms like “close contact” and “caress” exhibited relatively high alignment. “temptation” was frequently associated with curves, showing low variation but also low consistency. Our findings demonstrate that while AI can visualize abstract concepts, its effectiveness depends on prompts and motifs. Interpretation varies by individual sensitivity, but common aesthetic values reduce variation. This research highlights the need for educational methods accommodating diverse sensitivities and suggests new possibilities for AI-assisted Design education.