Immune checkpoint inhibitors (ICI) have shown positive results in patients with hepatocellular carcinoma (HCC). As liver function contributes to prognosis, its precise assessment is necessary for the safe prescribing and clinical development of ICI in HCC. We tested the accuracy of the albumin-bilirubin (ALBI) grade as an alternative prognostic biomarker to the Child-Turcotte-Pugh (CTP). In a prospectively maintained multi-centre dataset of HCC patients, we assessed safety and efficacy of ICI across varying levels of liver dysfunction described by CTP (A to C) and ALBI grade and evaluated uni- and multi-variable predictors of overall (OS) and post-immunotherapy survival (PIOS). We studied 341 patients treated with programmed-death pathway inhibitors (n = 290, 85%). Pre-treatment ALBI independently predicted for OS, with median OS of 22.5, 9.6, and 4.6 months across grades (p < 0.001). ALBI was superior to CTP in predicting 90-days mortality with area under the curve values of 0.65 (95% CI 0.57–0.74) versus 0.63 (95% CI 0.54–0.72). ALBI grade at ICI cessation independently predicted for PIOS (p < 0.001). Following adjustment for ICI regimen, neither ALBI nor CTP predicted for overall response rates or treatment-emerging adverse events (p > 0.05). ALBI grade identifies a subset of patients with prolonged survival prior to and after ICI therapy, lending itself as an optimal stratifying biomarker to optimise sequencing of systemic therapies in advanced HCC.