Congenital diaphragmatic hernia (CDH) is a life‐threatening condition with high morbidity and mortality rates. The survival rate of neonates with severe CDH is reportedly only 10%–15%. However, prenatal prediction of severe cases is difficult, and the discovery of new predictive markers is an urgent issue. In this study, we focused on microRNAs (miRNAs) in amniotic fluid‐derived small EVs (AF‐sEVs). We identified four miRNAs (hsa‐miR‐127‐3p, hsa‐miR‐363‐3p, hsa‐miR‐493‐5p, and hsa‐miR‐615‐3p) with AUC > 0.8 to classify good prognosis group and poor prognosis group in human study. The AUC for hsa‐miR‐127‐3p and hsa‐miR‐615‐3p, for predicting the poor prognosis, were 0.93 and 0.91, respectively. In addition, in the in vivo study, the miRNA profiles of the lung tissues of CDH rats were different from those of control rats. Additionally, two elevated miRNAs (rno‐miR‐215‐5p and rno‐miR‐148a‐3p) in the lung tissues of CDH rats were increased in the AF‐sEVs of CDH rats. Our results suggest that severe CDH neonates can be predicted prenatally with high accuracy using miRNAs contained in AF‐sEVs. Furthermore, miRNA profile changes in AF‐sEVs reflected the lung status in CDH. Our findings may contribute to the development of advanced perinatal care for patients with CDH.