Accessing complex online content remains a significant challenge for visually impaired users. This paper presents a Chrome extension that enhances web accessibility by providing real-time content extraction and summarization using Large Language Models (LLMs). Powered by the Llama model, the extension generates concise and contextually relevant summaries, displayed alongside selected webpage elements during keyboard-based navigation. Key features include dyslexia-friendly text, content saturation controls, and semantic evaluation to ensure accuracy and relevance. The system achieves efficient performance with an average processing time of 4.87 s (std. dev: 2.39) and demonstrates strong semantic preservation, reflected in a Cosine Similarity score of 0.989 and a low KL Divergence score of 0.025. Contextual accuracy is validated through a BERTScore of 0.808. While readability metrics, such as a FRE score of 27.46, highlight the complexity of some outputs, features like content customization improve usability. Comparative analysis demonstrates the system’s superiority over existing tools (ChatGPT, Tomedes, Notedly.ai) in semantic preservation (Cosine Similarity: 0.989 vs 0.985–0.986) while maintaining competitive readability. This work uniquely combines semantic accuracy with accessibility features, enhancing the web browsing experience for visually impaired users. The proposed solution contributes to more inclusive, AI-driven applications in web accessibility.