Lost in Instructions: Study of Blind Users' Experiences with DIY Manuals and AI-Rewritten Instructions for Assembly, Operation, and Troubleshooting of Tangible Products

A new study titled 'Lost in Instructions' reveals a critical gap in AI assistance for blind users performing DIY tasks with physical products. Researchers from Stony Brook University and other institutions conducted IRB-approved interviews and usability studies with 15 blind participants, examining how they use AI tools like ChatGPT and Be-My-AI alongside product manuals for assembly, operation, and troubleshooting.

The findings are stark: while product manuals are essential resources, their instructions are often inadequate for blind users who require spatial reasoning and structural understanding. More concerning, current AI tools frequently exacerbate these problems by generating incomplete, incoherent, or misleading guidance. The 28-page paper, submitted to CHI 2026, documents how AI fails to translate visual and spatial information into actionable, sequential steps that blind users can reliably follow.

This research highlights a significant accessibility challenge as AI becomes more integrated into daily life. The study's authors suggest specific improvements for AI systems, emphasizing the need for tailored instruction generation that accounts for the unique requirements of blind users interacting with tangible objects. The work underscores that current large language models lack the contextual understanding necessary for physical task guidance, pointing toward needed advancements in multimodal AI and specialized training data.