Digital accessibility ensures that websites, tools, and technologies are usable by people with disabilities, as outlined by the W3C. AI has increasingly contributed to improving digital accessibility through innovations like hand movement recognition for assistive devices and remediation tools such as ChatGPT, which resolves accessibility issues automatically. Additionally, accessibility overlays claim to enhance website functionality by integrating third-party code. However, challenges persist, including biases in AI algorithms, incomplete dataset representation, and limitations in context-specific issues, leading to user frustrations such as insufficient alt text descriptions. Concerns have also been raised about overreliance on accessibility overlays for achieving compliance with accessibility standards, which may overlook deeper issues. This research assesses AI’s role in improving digital accessibility using a mixed-method approach. Beyond exploring AI’s technical contributions, it critically engages with the Social Model of Disability and Disability Justice Model to examine how AI-driven solutions may either dismantle or reinforce societal barriers and systemic ableism. Through two case studies; reinforcement learning for web navigation and supervised learning for content classification, this research identifies both strengths and limitations of AI applications. Findings emphasize the need to address biases, improve data representation, and resolve technical challenges. Future work will focus on gathering stakeholder insights to develop strategic recommendations for more inclusive AI-driven systems that fully reflect the lived experiences of disabled communities.