This paper examines disability representation in images generated by text-to-image (T2I) generative AI systems. Through eight focus groups with 25 people with disabilities, the authors found that T2I models often present reductive archetypes for different disabilities, reflecting broader societal stereotypes and biases. Participants expressed concerns about the negative impacts of these representations, including perpetuating ableism and reinforcing biases in daily interactions. They offered suggestions for improving disability representation, such as showing multiple, heterogeneous images for a single prompt and including metadata to indicate the presence of disability. The study highlights the need for more diverse and accurate disability representation in AI-generated images and provides recommendations for improving T2I systems through interface and development changes. The findings also emphasize the importance of engaging with diverse communities to develop more respectful and inclusive AI systems.This paper examines disability representation in images generated by text-to-image (T2I) generative AI systems. Through eight focus groups with 25 people with disabilities, the authors found that T2I models often present reductive archetypes for different disabilities, reflecting broader societal stereotypes and biases. Participants expressed concerns about the negative impacts of these representations, including perpetuating ableism and reinforcing biases in daily interactions. They offered suggestions for improving disability representation, such as showing multiple, heterogeneous images for a single prompt and including metadata to indicate the presence of disability. The study highlights the need for more diverse and accurate disability representation in AI-generated images and provides recommendations for improving T2I systems through interface and development changes. The findings also emphasize the importance of engaging with diverse communities to develop more respectful and inclusive AI systems.