This study investigates the use cases of AI-powered scene description applications for blind and low vision (BLV) individuals. Through a two-week diary study involving 16 BLV participants, the researchers collected data on how participants used the application and their assessments of the visual descriptions provided. The study found that common use cases included identifying objects and their features, understanding surroundings, and reading text. Surprising goals included avoiding dangerous or disgusting objects and checking for others' presence. Participants rated the descriptions relatively low, with an average satisfaction score of 2.76 out of 5 and a trust score of 2.43 out of 4, indicating that significant improvements are needed to enhance user experiences. The study also explored the reasons why participants chose AI-powered applications over human assistants, highlighting factors such as independence, privacy, and the belief in AI's unbiased nature. The findings contribute to the development of more effective and trustworthy AI-powered scene description applications for BLV users.This study investigates the use cases of AI-powered scene description applications for blind and low vision (BLV) individuals. Through a two-week diary study involving 16 BLV participants, the researchers collected data on how participants used the application and their assessments of the visual descriptions provided. The study found that common use cases included identifying objects and their features, understanding surroundings, and reading text. Surprising goals included avoiding dangerous or disgusting objects and checking for others' presence. Participants rated the descriptions relatively low, with an average satisfaction score of 2.76 out of 5 and a trust score of 2.43 out of 4, indicating that significant improvements are needed to enhance user experiences. The study also explored the reasons why participants chose AI-powered applications over human assistants, highlighting factors such as independence, privacy, and the belief in AI's unbiased nature. The findings contribute to the development of more effective and trustworthy AI-powered scene description applications for BLV users.