Objective To systematically evaluate the pain prediction model for patients after transcatheter arterial chemoembolization，and to provide reference for selecting, optimizing and developing relevant predictive models. Methods PubMed, MEDLINE, CINAHL, Web of Science, Embase, CNKI，WanFang, VIP databases and SinoMed were searched for pain prediction models for patients after transcatheter arterial chemoembolization from the inception to 1 April 2025. Two researchers used the data extraction form for prediction model research and the bias risk assessment tool to extract data and evaluate the quality. Results A total of 15 papers and 21 predictive models were included, with a total sample size of 98-857 cases, the number of positive events ranges from 27 to 206, the area under the curve(AUC)of the model was 0.676 to 0.916. Of these, 8 models were reported to be calibrated, 6 were externally validated,predictors of the multivariate model included age, history of abdominal pain after transcatheter arterial chemoembohzafion,history of transcatheter arterial chemoembohzafion treatment, distance from the liver capsule, operation mode, tumor size, and number of embolizedtumors. All studies demonstrated good overall applicability, but all had a high overall risk of bias. Conclusion The current pain prediction models for patients after transcatheter arterial chemoembolization still have limitations. In the future, high-performance and low-bias risk pain prediction models should be developed to guide clinical practice.