Cognitive Accessibility in Generative AI Interfaces: A Systematic Review

Generative Artificial Intelligence (GAI) represents a significant shift in the way people interact with technology, with a growing impact in sectors such as education, communication, and content creation. Although its applications promise relevant benefits for people with cognitive disabilities, such as language simplification or support for task organization, there are also significant accessibility challenges that remain unexplored. This paper presents the results of a systematic literature review focused on the cognitive accessibility of textual GAI interfaces. The main objective was to identify the current cognitive accessibility challenges inherent in textual GAI interfaces and to synthesize the solutions proposed to promote a more inclusive interaction. To this end, seven studies selected between 2020 and 2024 from four academic databases (Scopus, Web of Science, IEEE Xplore, and ACM Digital Library) were analyzed. The analysis reveals that current interfaces impose excessive cognitive load and suffer from systemic lack of predictability and user scaffolding due to their reliance on complex prompt engineering and low explainability. Emerging proposals based on inclusive design, personalization, multimodal interfaces, and ethical frameworks are also identified. The paper concludes by highlighting the need for more empirical studies, specific design guidelines, and validated evaluation methods to ensure that generative technologies are truly accessible to all.