This research evaluates how well Multimodal Large Language Models (MLLMs) serve as visual assistants for visually impaired individuals, identifying both capabilities and critical gaps.

• High adoption rate among visually impaired users despite limitations
• Key challenges include contextual understanding, cultural sensitivity, and complex scene interpretation
• Models often struggle with the specific informational needs of visually impaired users
• Findings highlight the need for specialized development focusing on accessibility requirements

For support professionals, this research provides crucial insights into how to better design and implement AI-based visual assistance technologies that truly meet the needs of visually impaired users rather than relying on general-purpose models.

Evaluating Multimodal Language Models as Visual Assistants for Visually Impaired Users