The study, written by Psilopatis et al., examines the accuracy of various large language model tools (LLMs) in Cardiotocography (CTG) interpretations with regard to the Federation of Gynecology and Obstetrics criteria (FIGO). This paper analyzed sixty CTG traces, as classified by University Hospital Basel, by presenting screenshots to four different LLMs. The LLMs (Chat-GPT-4.0, Google Gemini, Bing CoPilot, and DeepSeek) would classify them as either normal or abnormal. Specifically, the LLM that provided a satisfactory interpretation of normal CTGs was then tasked to classify 30 abnormal CTG traces. The observed results were that DeepSeek was unable to evaluate any CTG traces due to the LLM’s lack of features to process visuals. Google Gemini inaccurately classified 93.3% of the CTG traces, most likely due to variable decelerations despite being considered normal fluctuations. Chat-GPT-4.0 inaccurately classified 53.3% CTG traces as being suspicious again due to variable decelerations. Finally, Bing Copilot correctly identified 96.6% CTG traces as being normal, presenting a high alignment with obstetrician-based classifications, but failed to identify the abnormal CTG traces. These findings indicate that general-purpose LLMs face substantial limitations in accurately evaluating CTG traces based on FIGO criteria, with DeepSeek being the only one of four showing some promise, though still unable to identify CTG data entirely.