This study analyzed the consistency of AI-generated product descriptions for New Hanbok images to examine the feasibility of applying generative AI to product pages in online fashion shopping malls. A total of 2,004 product and styling images were collected, and product descriptions were generated using ChatGPT. Document similarity analysis was conducted to evaluate the consistency of the generated descriptions. The result showed that the AI tool maintained a generally consistent interpretation of New Hanbok product designs, although the level of consistency varied depending on item and the styling method. In comparisons between product images and styling images, dresses and coats showed relatively high similarity, whereas jackets and T-shirts showed lower similarity. When an item primarily determined the overall design and was less affected by coordination with other garments, the AI tool consistently generated similar descriptions across images. In contrast, for items whose appearance and impression were more strongly influenced by styling combinations, it tended to produce more varied descriptions reflecting the overall styling context. Comparisons among styling images revealed relatively high similarity for skirts and jackets, while coats and dresses exhibited greater variation depending on styling. Overall, the tool tended to give products with similar design characteristics and styling directions consistent descriptions across different images. These findings demonstrate the potential of generative AI to support the drafting of product descriptions for online fashion shopping malls. The study also provides foundational evidence for the development of AI-assisted product detail pages and effective prompt design strategies for fashion e-commerce.

Abstract
I. Introduction
Ⅱ. Theoretical Background
    1. Online shopping mall product detail pages
    2. The fashion industry and generative artificialintelligence
Ⅲ. Research Method
Ⅳ. Results and Discussions
    1. Similarity analysis results for productdescriptions of product images (PI) and stylingimages (SI)
    2. Similarity analysis results for productdescriptions among styling images (SIs)
Ⅴ. Conclusion
References

• Hee Young Kim(Assistant Professor, Dept. of Fashion Design, Kyungin Women’s University, Korea) | 김희령 (경인여자대학교 패션디자인학과 조교수) Corresponding author