This chapter critically examines the nuanced challenges of disability representation in AI-generated content, with a specific focus on OpenAI’s ChatGPT-4. In the first section, we summarize a critical content analysis of 50 short stories about disabled children generated by ChatGPT-4 in August 2023. Based on our analysis, we offer the concept of disability evasiveness, a phenomena whereby non-disabled people claim to ‘not see’ disability. We describe three facets of disability evasiveness that are evident in the stories and speculate on the technical mechanisms and bias mitigation tools that may have cause them. In the second section, we extend the original study in two ways. First, we generate new stories and explore how the stories have changed over the year and a half since the original study as a result of OpenAI’s ongoing bias mitigation tools and system updates. Second, we explore the impact of three different prompts: one using person-first language, one using identity-first language, and a highly engineered prompt. Our analysis not only underscores the complexities of AI-mediated representations but also emphasizes the ongoing challenges in mitigating biases and achieving equitable portrayals in digital content creation. The chapter concludes with a discussion on what we want from disability representation in AI-generated literature.