Liver biopsy is currently the only reliable method to establish nonalcoholic fatty liver disease (NAFLD) severity. However, this technique is invasive and occasionally associated with severe complications. Thus, non-invasive diagnostic markers for NAFLD are needed. Former studies have postulated 18 different serum microRNA biomarkers with altered levels in NAFLD patients. In the present study, we have re-examined the predictive value of these serum microRNAs and found that 9 of them (miR-34a, -192, -27b, -122, -22, -21, -197, -30c and -16) associated to NAFLD severity in our independent cohort. Moreover, miR-192, -27b, -22, -197 and -30c appeared specific for NAFLD, when compared with patients with drug-induced liver injury. Preliminary serum RNAseq analysis allowed identifying novel potential miRNA biomarkers for nonalcoholic steatohepatitis (NASH). The classification performance of validated miRNAs (and their ratios) for NASH was better than that reached by AST, whereas for advanced fibrosis prediction miRNAs did not perform better than the FIB-4 algorithm. Cross-validated models combining both clinical and miRNA variables showed enhanced predictivity. In conclusion, the circulating microRNAs validated demonstrate a better diagnostic potential than conventional serum markers to identify NASH patients and could complement and improve current fibrosis prediction algorithms. The research in this field is still open.