Enhancing Readability of Surgical Patient Education Resources Using ChatGPT

Patient education materials are essential for informed decision-making. However, their readability often exceeds the sixth-grade reading level recommended by the American Medical Association, creating barriers for patients with limited literacy. Artificial intelligence (AI) tools such as ChatGPT-4o may offer a scalable approach to improving accessibility.

We analyzed 34 publicly available online patient education materials across general surgery, neurosurgery, orthopedic surgery, and plastic surgery. ChatGPT-4o was prompted to simplify the text to the sixth-grade reading level. Two independent surgical experts reviewed text for accuracy and comprehensiveness. The primary outcome was the Simple Measure of Gobbledygook index. Secondary outcomes included Coleman-Liau index, Flesch-Kincaid grade level, word count, and reading time. Accuracy was measured by the number of inaccurate sentences, and comprehensiveness by coverage of six content categories.

ChatGPT-4o simplification significantly decreased mean Simple Measure of Gobbledygook score from 12.1 to 10.5 (P < 10^-9), with similar reductions in Coleman-Liau index (-0.8, P < 10^-5) and Flesch-Kincaid (-1.6, P < 10^-8). Word count reduced by 55% on average, most notably in plastic surgery (66%, P = 0.008). Accuracy was maintained in all specialties; however, key content such as risks and benefits was often omitted.

AI can improve the readability of surgical patient education materials while maintaining accuracy. Achieving the recommended sixth grade reading level may be challenging due to the inherent complexity of medical information. A hybrid approach that combines AI-assisted simplification with clinician review may represent the most effective strategy for efficiently developing patient-facing educational resources that are accessible, accurate, and comprehensive enough to support informed decision-making.