Abstract
<title>Abstract</title> <p>Purpose Olfactory dysfunction (OD) affects approximately 22% of the population and significantly impairs quality of life. Patients increasingly seek health information online, including AI chatbots, yet its quality for OD remains poorly characterised. This study evaluated the quality and readability of OD information across five digital health platforms using validated instruments. Methods Five platforms were assessed: Google Search, ChatGPT (GPT-5.3), Gemini 3, Claude Sonnet 4.6, and Perplexity AI search. Eight standardised search terms spanning medical and lay terminology were applied, and responses were independently assessed by two raters using DISCERN, QAMAI (AI platforms only), and readability metrics (Flesch-Kincaid Grade Level; SMOG Index). Inter-rater reliability was calculated using Cohen's kappa and intraclass correlation coefficient (ICC). Results Perplexity achieved the highest mean DISCERN score (50.56 ± 5.74), with all sources rated moderate quality. Google Search produced the only high-quality sources (3.75%) and showed a high variability (41.96 ± 10.14). ChatGPT (43.00 ± 6.27) and Gemini (44.94 ± 12.05) had similar mean DISCERN scores, although Gemini was more variable. Claude scored lowest (38.13 ± 4.57), with 75% of sources rated low quality. Reading difficulty correlated weakly with information quality (Spearman's ρ = 0.255, p = 0.0067). Mean readability (FKGL 11.01) exceeded NHS recommendations (Grades 6–8), with only 27.7% of sources meeting the target. Conclusion Google alone produced high-quality OD information by DISCERN criteria; AI-generated content consistently failed to meet this standard, particularly for treatment-related domains. Readability was poor across all platforms. Clinicians should direct patients to validated specialty resources and highlight AI limitations, particularly regarding treatment risks, non-treatment consequences, and quality-of-life impact.</p>