Objectives To evaluate the accuracy and reproducibility of semi-quantitative visual assessment of hepatic tumor burden (HTB) on MRI and to investigate its prognostic value in predicting overall survival (OS) in neuroendocrine liver metastases (NELMs) treated with transarterial chemoembolization (TACE). Methods Three independent readers blinded to the quantitative HTB measurement reviewed baseline MRI in 111 NELM patients treated with TACE. Readers visually assessed and semi-quantitatively categorized HTB using the European Neuroendocrine Tumor Society (ENETS) guidelines. Quantitative HTB measured by manual segmentation was used as the reference standard. Agreements between quantitative and semi-quantitative measurement of HTB, as well as intra- and inter-reader reproducibility, were evaluated using weighted kappa coefficient and intraclass correlation coefficient (ICC). Survival analysis included the Kaplan-Meier curves and Cox regression. Harrell C-index was calculated to evaluate the prognostic value of semi-quantitative HTB for predicting OS. Results According to quantitative HTB, 41, 29, 25, and 16 patients were categorized into <= 10%, 11-25%, 26-50%, and > 50% groups, respectively. Agreements between quantitative and semi-quantitative measurement of HTB by each reader (weighted kappa, 0.82-0.96), intra-reader agreement (weighted kappa, 0.95), and inter-reader agreements (weighted kappa, 0.84-0.91; ICC, 0.98) were at least substantial to almost perfect. Semi-quantitative HTB was an independent prognostic factor in NELMs treated with TACE (multivariate Cox regression, p < 0.001), with prognostic value comparable to that of quantitative HTB (Harrell C-index, 0.735 for both semi-quantitative and quantitative HTB in multivariate regression). Conclusion Semi-quantitative visual assessment of HTB using MRI is accurate and reproducible and could reliably predict OS in NELMs treated with TACE.