Readability assessment of ChatGPT-4 and ChatGPT-5 responses to ear care questions

Abstract

Aim: To assess the readability of ChatGPT-4o and ChatGPT-5 responses to questions about ear care. Materials and Methods: Ten questions about ear care from the World Health Organization’s ‘Questions and Answers’ section were each posed seven times to both ChatGPT-4o and ChatGPT-5. A total of 70 responses were obtained and analyzed for readability. Results: For ChatGPT-4o, the mean readability scores were: Flesch Reading Ease 62.84 ± 8.32, Flesch–Kincaid Grade Level 8.13 ± 1.59, Simple Measure of Gobbledygook 7.35 ± 1.42, Gunning Fog Index 8.44 ± 1.79, Coleman–Liau Index 9.62 ± 1.71, and Automated Readability Index 8.70 ± 2.02. For ChatGPT-5, the corresponding scores were: Flesch Reading Ease 72.93 ± 7.69, Flesch–Kincaid Grade Level 5.75 ± 1.17, Simple Measure of Gobbledygook 6.35 ± 0.79, Gunning Fog Index 7.37 ± 1.22, Coleman–Liau Index 7.32 ± 1.52, and Automated Readability Index 5.78 ± 1.42. ChatGPT-5 demonstrated better readability than ChatGPT-4o across all readability indices (p<0.001). Conclusion: The readability of the answers produced by ChatGPT-4o is “standard”, and that of ChatGPT-5 is “fairly easy”. ChatGPT-5 responses showed improved readability over ChatGPT-4o, aligning more closely with patient education readability standards and underscoring its potential as an accessible source of health information.