Volume 2 - Issue 2, March - April 2026

The rapid growth of web-based information systems has increased the cognitive and environmental cost of accessing digital content, particularly for users with diverse accessibility needs. Many existing AI-driven accessibility solutions rely on cloud-based, computationally intensive models that raise concerns related to latency, privacy, and energy consumption. In this paper, we propose an adaptive, explainable, and lightweight AI-driven framework for real-time web content accessibility that aligns with sustainability principles. The framework dynamically adjusts textual complexity and presentation using interpretable linguistic analysis and implicit user behaviour, while operating entirely on the client side to minimize computational overhead and carbon footprint. We introduce three core components: (i) an Adaptive Readability Optimization (ARO) algorithm for real-time text simplification, (ii) a Behaviour-Aware ersonalization Model (BAPM) that incrementally learns user preferences using implicit interaction signals, and (iii) an Explainable Difficulty Classification Framework (EDCF) for transparent content difficulty assessment. Experimental evaluation demonstrates significant improvements in readability, low processing latency, and energy-efficient operation suitable for sustainable AI deployment. The proposed approach highlights how lightweight, explainable AI systems can contribute to inclusive and sustainable digital ecosystems.

Tanusri P, Gobi I,"An Adaptive, Explainable and Lightweight Framework for Real-Time Web Content Accessibility" International Journal of Advanced Multidisciplinary Research and Educational Development, V2(2): Page(940-943) Mar-Apr 2026. ISSN: 3107-6513. www.ijamred.com. Published by Scientific and Academic Research Publishing.