Addressing Web Accessibility Challenges with Generative Artificial Intelligence Tools for Inclusive Education

This article addresses the comprehensive evaluation of web accessibility in 20 generative artificial intelligence (AI) applications, such as ChatGPT and DALL-E, through a five-phase approach. Common issues were identified, including inadequate image descriptions, lack of semantic structures, and keyboard navigation challenges. Despite the inherent complexity of generative tools, the importance of evaluating their accessibility to ensure the inclusion of users with diverse abilities is emphasized. The WAVE automatic tool was used to identify issues, and future directions are proposed, such as improving image descriptions and optimizing keyboard navigation. The most significant accessibility challenges are linked to minimal contrast, representing 38%, followed by issues in easy-to-read font and text alternatives, both at 15%, associated with the perceptible principle. The discussion covers areas for improvement, ethical implications, and strategies to foster continuous enhancement in generative AI accessibility, highlighting the importance of balancing benefits and ethical challenges. Future research includes developing techniques for automatically generating consistent image descriptions, refining semantic structures, and optimizing keyboard navigation. Additionally, it is imperative to improve automatic accessibility evaluation tools to address the unique challenges of generative AI applications.