Double-blind pilot study reveals AI improves surgical consent comprehensibility over standard forms, suggesting potential for hybrid expert workflows.
Key Points
To evaluate whether large language models (ChatGPT-5.2 and Gemini 3 Pro) can produce more comprehensible otolaryngology surgical consent forms compared to standard official documents.
Prospective, double-blind design assessing 123 Turkish consent documents spanning 41 otolaryngology surgical procedures across three sources: standard official forms, ChatGPT-5.2, and Gemini 3 Pro.
Three independent otolaryngologists evaluated medical quality across five 5-point criteria, five blinded lay volunteers rated comprehensibility, and readability was measured using the Ateşman and Bezirci-Yılmaz indices.
Group comparisons and effect-size parameter estimations were evaluated using linear mixed-effects models.
Large language models, particularly Gemini 3 Pro, substantially enhanced the comprehensibility and readability of consent documents relative to standard forms.
Improvements in readability occurred alongside slight reductions in medical precision and ethical-legal adequacy, highlighting the necessity of clinical oversight.