The study by Engelstein et al. aims to determine if the Generative Pretrained Transformer 4 Omni (GPT-4o) is effective in recognizing and classifying abnormalities in 12-lead ECGs. Using 80 randomly selected ECGs (including 30 normal ECG strips and 50 ECG stips with varying diagnoses), OpenAI’s GPT-4o was prompted to first recognize an ECG image, classify it as normal or abnormal, and classify the image to 1 of 6 clinical diagnoses which were already pre-evaluated. In the first scenario, ECG image identification, the GPT-4o model was able to correctly identify and classify 100% of the images as ECGs. In the second scenario, classifying normal abnormal ECGs, the GPT-4o model was not trained and yielded an accuracy between 53% to 63%. After being trained and given an example of evaluation criteria, the accuracy increased to an average of 83%. The last scenario, identifying specific clinical diagnosis, achieved an accuracy of 28% and 41% (without training and with training). Due to the model’s success in ECG image identification, it is interpreted that GPT-4o may be most helpful as a supplementary tool and may be used for early and simple interpretation.