AI-Based Web Accessibility Framework for Assisting Deaf and Visually Impaired Users

Despite two decades of accessibility standards, the digital world remains largely unnavigable for millions with 
sensory disabilities, a problem defined not just by a lack of access, but a lack of meaningful usability—a "second-level 
digital divide." Existing assistive technologies are often rigid, placing a significant cognitive burden on users to 
interpret poorly structured content. This paper introduces the Dynamic Accessibility Transformation Engine 
(DATE), a novel, AI-driven framework that actively re-engineers web content into accessible, multi-modal formats. 
Our methodology combines a constructivist grounded theory analysis of user needs with a human-centered AI 
design process. The developed prototype, an intelligent chatbot integrated into a web application, addresses the 
critical "semantic gap" in current AI by not just recognizing content but interpreting its context to provide 
synthesized outputs like contextual image descriptions, thematic summaries, and synchronized audio-visual 
transcriptions. A mixed-methods usability study was conducted with 10 participants (5 visually impaired, 5 deaf). 
The results show that the DATE prototype significantly reduced cognitive load, as measured by the NASA-TLX index, 
and increased task success rates. Thematic analysis of qualitative data revealed themes of user empowerment and 
reduced frustration. This research contributes a new model for accessibility—one that shifts the burden of 
adaptation from the user to a proactive, context-aware computational system, offering a vital step toward a more 
equitable socio-technical future.