This study aimed to analyze answers provided by Wenxin Yiyan and 360 Intelligence Brain to questions in evidence-based medicine in terms of correct rates and potential causes, in order to provide evidence in the learning/teaching activities with AI. Totally 513 questions were included, with the answers given by the teachers serving as gold standard, and correct rates of the two AI APPs were calculated. Pearson Chi-Squared test and McNemar test were used to analyze the answers for single-choice questions, multiple-choice questions, and all the questions between the two groups. Possible reasons for the wrong answers were also analyzed. It was found that the total correct rates of Wenxin Yiyan and 360 Intelligence Brain were 76.4% (95%CI [72.5%, 80.0%]) and 89.5% (95%CI [86.5%, 92.0%]) (χ2 = 30.926, P = 2.680E-08), respectively. The correct rates of single- and multiple-choice questions of Wenxin Yiyan were respectively 82.9% (95%CI [78.7%, 86.6%]) and 58.7% (95%CI [50.0%, 67.0%]) (χ2 = 32.881, P = 9.800E-09). The correct rates of single- and multiple-choice questions of 360 Intelligence Brain were 93.3% (95%CI [90.3%, 95.6%]) and 79.0% (95%CI [71.2%, 85.5%]) (χ2 = 22.049, P = 2.657E-06). The differences were statistically significant. In addition, the correct rate of answers given by AIs to single-choice questions was higher than that of answers to multiple-choice questions. Our findings indicate that the correct rates of 360 Intelligence Brain are higher than those of Wenxin Yiyan either for single- or multiple-choice questions. The correct rates of single-choice questions in both AI-APPs are higher than those of multiple-choice questions in evidence-based medicine homework and examination.