To determine the diagnostic performance and inter-reader agreement of the contrast-enhanced ultrasound liver imaging reporting and data system (CEUS-LI-RADS) for diagnosing hepatocellular carcinoma (HCC) in high-risk patients. In this prospective study, CEUS-LI-RADS categories (LR-5 for predicting HCC) were assigned by six blinded readers and compared to the definitive HCC diagnosis in patients with liver cirrhosis per the 2017 China Liver Cancer Guidelines (CLCG). CEUS features were recorded in 96 patients with 96histology-proven lesions. The diagnostic performance of LR-5 was described by the sensitivity, specificity and accuracy. Multi-reader agreement was assessed by using intraclass correlation coefficients (ICC). In cirrhotic patients, the specificity of LR-5 (range: 92.7-100.0 %) was statistically higher than that of CLCG for each reader (range: 28.6-64.3 %). However, the sensitivity (range: 38.6-63.6 %) and accuracy (range: 53.4-70.7 %) were statistically lower in CEUS-LIRADS than in CLCG (sensitivity range: 88.6-100.0 %; accuracy range: 77.6-86.2 %). Only fair to moderate inter-reader agreement was achieved for the CEUS-LI-RADS category (ICC = 0.595) and washout appearance (ICC range: 0.338 to 0.555). Neither nodule-in-nodule nor mosaic architecture was observed more often in HCC (all P > 0.05), with poor inter-reader consistency for both (both ICC < 0.20). CEUS-LI-RADS category 5 has a high specificity but a low accuracy for identifying HCC in high-risk patients. Inter-reader agreement is not satisfactory concerning CEUS-LIRADS category and washout appearance. Moreover, the clinical value of ancillary features favoring HCC is quite limited.