Empowering Blind and Low-Vision Learners through an Agentic AI Journalist: A User-Centered Accessibility Framework for Real-Time Global News

Access to timely and trustworthy news remains a major challenge for blind and low-vision learners in low- and middle-incomecountries (LMICs), where most digital platforms lack screen-reader compatibility and efficient non-visual navigation. This studyintroduces the Agentic AI Journalist, a lightweight, voice-first assistant designed to deliver real-time, verifiable briefings optimized forauditory interaction. The system combines a FastAPI backend with Model Context Protocol (MCP) orchestration to coordinate webretrieval, multi-source verification, and prosody-aware summarization. Evaluation with blind students in Bangladesh measured bothquantitative performance—factual accuracy, latency, and usability—and qualitative outcomes, including trust, comprehension, andcognitive effort. Results show a median time-to-first-audio of 1.5 seconds on entry-level hardware, strong factual correspondence (F1 =0.92), and high perceived usefulness for academic and civic information access. Participants reported enhanced trust and skimmabilitycompared to generic text-to-speech systems. These findings demonstrate that agentic verification and prosody-first design can makereal-time news access more inclusive, resource-efficient, and socially empowering for blind learners in resource-constrained settings.