Objective <p>Gastric cancer liver metastasis (GCLM) has become one of the major contributors to gastric cancer (GC) fatality. It is essential to assess the risk of liver metastasis in GC patients so that surgeries can be scheduled promptly. The nomogram-based prediction model has been well exploited in recent studies to evaluate overall survival in various cancers, including GC. Still, such a model was rarely proposed for predicting GCLM risk. This study was aimed at developing and validating a statistical model to predict synchronous liver metastases at diagnosis in patients with gastric cancer (GC) using variables from the SEER database.</p> Methods <p>The retrospective study uses data extracted from the SEER database. A total of 7787 patients were divided into a training set (<i>n</i> = 5444) and a validation set (<i>n</i> = 2343).</p> Results <p>A univariate and a multivariate analysis were carried out to screen out seven characteristics significantly related to the GCLM outcome, namely age, sex, T stage, N stage, tumor size, surg-prim-site, and radiotherapy (<i>P</i> &lt; 0.05), and a nomogram was subsequently constructed. The receiver operating characteristic curve (ROC) indicated a good discriminative ability of the model. The area under the curve (AUC) of the training cohort and validation cohort was 0.858 (95% CI:0.846–0.870) and 0.849 (95% CI:0.831–0.868), respectively. The calibration curve showed a good agreement between the prediction and observed outcomes in both data sets. The decision curve analysis in the training and validation cohorts showed that the model had a good net clinical benefit.</p> Conclusion <p>The statistical performance demonstrated that the model performed well in predicting GCLM risk and its capability to support treatment regimen adjustment.</p>