Large Language Models as Patient Education Tools in Hypothyroidism: A Cross-Sectional Analysis of Dietary, Pharmacological, and Safety Recommendations

Read full text
Authors: Patryk Hebda, Mateusz Kubicki, Andrii Bilyk, Maria C. Faber Krol, Przemysław Kołodziej, Roman Cemaga, Adam Wolski, Natalia Marta Bruska, Ewa Szplit, Katarzyna Więckowska
Cited by: 0
References: 23

Abstract

Background. Hashimoto’s thyroiditis (HT) is the most common cause of hypothyroidism worldwide and requires lifelong management. Due to the chronic nature of the disease and persistent symptoms despite biochemical euthyroidism, patients frequently seek online advice on dietary restrictions, supplementation, and pharmacotherapy. Large Language Models (LLMs) such as ChatGPT are increasingly used as sources of patient education; however, their consistency with established clinical guidelines, particularly those relevant to European and Polish populations, remains insufficiently evaluated. Aim. This study aimed to conduct a comprehensive cross-sectional analysis of the accuracy, safety, and concordance of ChatGPT-4o responses with international clinical guidelines and clinical practice across four domains: diet/lifestyle, pharmacotherapy, supplementation, and common myths/safety. Materials and Methods., Thirty-five standardized patient prompts reflecting real-world clinical questions were analyzed. Responses generated by ChatGPT-4o (January 2026) were assessed using a 3-point concordance scale based on recommendations from the Polish Endocrine Society (PTE), European Thyroid Association (ETA), and relevant peer-reviewed literature. Results. Overall concordance with clinical guidelines was high (91.4%). The strongest performance was observed in pharmacotherapy-related questions and safety-critical scenarios, including pregnancy and drug interactions. Minor limitations concerned dosing nuances for certain supplements, where the AI occasionally adopted a more optimistic approach than conservative clinical recommendations. Conclusions. ChatGPT shows high potential as a supportive educational tool for patients with hypothyroidism, providing guidance largely consistent with Evidence-Based Medicine and current PTE/ETA standards. However, due to limited nuance in “grey area” topics, it should complement rather than replace professional endocrinological consultation.

Fields of Study & Subjects

Artificial Intelligence in Healthcare and Education Thyroid Disorders and Treatments Health Literacy and Information Accessibility

Publication details

Journal / source
Journal of Education Health and Sport
Volume
88
Pages
68691–68691
Publication date
8 March 2026
Language
EN