INTRODUCTION: Supraventricular tachycardia (SVT) is the most prevalent arrhythmia among young adults. With the rapid advancement of artificial intelligence technologies, natural language processing models (NLPMs) such as ChatGPT, Gemini, and Bing Chat are becoming increasingly widespread in the field of medicine. We aim to assess the precision and consistency of responses produced by ChatGPT-4o, Gemini, and Bing Chat to frequently asked questions regarding SVT.
METHODS: A list of fifty commonly asked questions regarding SVT was inquired twice, with a one-week interval, to ChatGPT-4o, Gemini, and Bing Chat. Two cardiologists assessed the responses from each NLPM without knowledge of each other’s evaluations. The content was rated using the following scale: totally correct (1), incomplete (2), and incorrect (3).
RESULTS: Most of the responses from all models were rated as either ‘totally correct’, ‘incomplete’, or ‘incorrect’. Even though ChatGPT-4o did not generate any ‘incorrect’ answers, Bing Chat and Gemini produced some incorrect responses. Regarding the accuracy of responses, ChatGPT achieved a score of 92%, Gemini obtained 70%, and Bing Chat reached 58%. ChatGPT-4o also achieved the highest ‘reproducibility’ score at 90%, followed by Gemini at 86%, and Bing Chat at 72%.
DISCUSSION AND CONCLUSION: Our study highlighted that ChatGPT-4o is capable of generating valuable answers to patients’ questions related to SVT. As NLPMs—especially ChatGPT-4o—continue to improve, they hold great potential for the management of chronic conditions like SVT.