Previous research has utilized user reviews and automatically generated image descriptions to enhance the accessibility of the online shopping experience. For example, one system gathered customer reviews of products to improve interactive information retrieval. However, variations in information accuracy can lead to misunderstandings. Furthermore, users who are blind or visually impaired (BLV) often seek external validation to supplement algorithmic outputs, wanting to understand how clothing looks on other individuals. These unresolved challenges hinder the overall accessibility of online shopping platforms. 

Generative Artificial Intelligence (GenAI) offers the possibility of more consistent and personalized user experiences, potentially simplifying information dissemination and catering specifically to the nuanced needs of the BLV community. Virtual try-on technologies present an opportunity for efficient and realistic 3D garment visualization. However, they remain inaccessible to BLV users because they primarily depend on visual feedback, which those with visual impairments cannot perceive, and they often lack auditory interfaces that could make them usable for the BLV community. This underscores the need for the development of inclusive technologies that address the unique challenges faced by the BLV community in online shopping.

Recent studies have highlighted the potential of GenAI models to enhance user experiences in virtual reality, demonstrating the powerful generation capabilities of these models. As a result, we aimed to explore the following research questions:

- What are BLV users' current online clothing shopping experiences and preferences regarding existing assistive technologies?
- How can GenAI enhance BLV users' online clothing shopping experiences?

To address these questions, we initially conducted a formative study to understand the current online clothing shopping experiences and challenges faced by the BLV community. Based on the findings, we identified two key design considerations and developed TailoredVision, a system designed to improve the shopping experience for BLV users. This system integrates a GenAI assistant with human crowdsourcing feedback, processing full-body images and clothing selections using virtual try-on technology. It produces integrated visualizations that are evaluated for style, occasion suitability, and weather appropriateness, closely aligning with user preferences. The incorporation of human crowdsourcing provides objective feedback that fosters independence and confidence within the BLV community.

Our preliminary research, which included a survey with 18 participants and focus groups with 9 individuals, aimed to gather insights from users. Additionally, we performed a preliminary user study with 10 BLV participants to assess the usability of TailoredVision. Results indicated that using the system significantly improved the task success rate for these participants when completing clothing purchasing tasks on e-commerce platforms, demonstrating its effectiveness in facilitating autonomous shopping for visually impaired users. Furthermore, feedback collected from user experience questionnaires and open interviews pointed to a positive user experience, emphasizing the system's potential to enhance the inclusivity and accessibility of online shopping for the BLV community.