Assessing Accessibility by GenAI. Using NotebookLM to Evaluate DUXAIT-NG for Accessibility Issues and Recommendations

Ledesma C.; Falconi F.; Moquillaza A. · Lecture Notes in Computer Science (2026) · DOI: 10.1007/978-3-032-30089-8_7

Resumen

The accelerated evolution of User Interfaces gives rise to a critical problem that, if not adequately managed, can negatively impact the very User Experience it initially seeks to optimize: Accessibility Regression. This study addresses this issue by implementing a GenAI-assisted Accessibility Evaluation Method that employs a GenAI tool based on an LLM (Google NotebookLM) as both an inference engine and an evidence-based knowledge management environment. This research pursues a dual objective. The first objective is to present a proposal for regulatory cross-referencing and to generate high-fidelity, high-quality qualitative insights applicable to Accessibility evaluation in Human–Computer Interaction. The second objective is to apply this method to a real-world case study in which a multivariate accessibility diagnosis of a platform was conducted, to integrate the resulting findings into the platform’s own evolution and development cycles. Results were promising. On the one hand, these findings confirm several hypotheses formulated during the application’s design phase; on the other hand, they reveal previously undetected accessibility issues. Representative examples of these findings include cognitive accessibility considerations, such as the need to simplify the flow and reduce mental workload throughout the user journey, as well as barriers affecting sensory and motor accessibility. In this context, the capacity of the proposed method and its supporting tool, NotebookLM, to automatically correlate empirical evidence, such as usage flows, with regulatory and design artifacts, demonstrates an advantage over the inherent limitations of conventional accessibility evaluation approaches. This advantage is particularly evident in optimizing the time and resources required to conduct comprehensive accessibility assessments. © The Author(s), under exclusive license to Springer Nature Switzerland AG 2026.

Instituciones: Pontificia Universidad Católica del Perú; Universidad Privada del Norte