Greek embroidered dowry textiles embody women’s craftsmanship, local knowledge, and familial memory. Although they are commonly preserved as cultural heritage objects, their responsible reinterpretation in contemporary fashion remains underexplored, particularly in relation to the emerging use of artificial intelligence in design. In this study, we examine how nine Greek embroidered dowry textiles can inform the design of contemporary garments through a practice-based design methodology. We explore the potential to reactivate the patterns, structures, and cultural associations of these textiles through reuse and reinterpretation, rather than treating them solely as preserved cultural heritage objects. The study addresses a realistic dilemma faced by designers working with sensitive historical textiles, regarding how to modify them without eliminating their visual identity or cultural significance, especially by integrating artificial intelligence tools into the design workflow. We adopted a research-through-design approach to develop and test a repeatable workflow for each textile. This included capturing patterns, analyzing visual elements, and deriving design rules. ChatGPT was used to develop ideas by experimenting with images, encouraging interactions between text and image, and providing feedback on iteration, and, in some cases, physical garment production. The results ranged from strong resemblance to the original embroideries to looser reinterpretations, retaining only certain elements. These findings were categorized into a fidelity-deviation taxonomy and translated into actionable guidelines for users. Designers aimed to harmonize similarity, innovative transformation, and cultural legibility in heritage-inspired fashion design.