This study examines how blind and low vision (BLV) individuals use and experience visual interpretation applications powered by multimodal large language models (MLLMs).

• BLV users increasingly rely on AI-powered tools for daily visual interpretation needs
• MLLMs offer more descriptive and contextual visual interpretations compared to previous technologies
• Users show growing trust in these applications, including for critical tasks like medication identification
• Research reveals both the promise and potential risks of MLLM adoption in assistive technology

Medical Impact: The findings highlight important safety considerations as BLV individuals increasingly trust these applications for medication identification and dosage interpretation, suggesting a need for improved reliability and appropriate user guidance in medical contexts.

Original Paper: Towards Understanding the Use of MLLM-Enabled Applications for Visual Interpretation by Blind and Low Vision People