We aimed to compare the early responder rates, defined as complete or partial responders, using response evaluation criteria in solid tumors (RECIST) 1.1, modified RECIST (mRECIST), and Choi criteria in advanced HCC patients treated with atezolizumab-bevacizumab (atezo-bev), and to correlate them with progression-free survival (PFS) and overall survival (OS). This retrospective study included advanced HCC patients treated with ≥ 3 cycles of atezo-bev. Two reviewers assessed responses using RECIST 1.1, mRECIST, and Choi criteria at 1st follow-up imaging. Kaplan-Meier curves with log-rank tests evaluated and compared PFS and OS. Cox proportional hazard models identified survival outcome predictors. Kappa statistics assessed inter-reader agreement. We evaluated 77 patients (65 men; mean age, 62.8 ± 12.3 years). Choi's criteria revealed the highest early responders rate (53.2%), exceeding mRECIST (32.5-33.8%) and RECIST 1.1 (24.7-26.0%), with an excellent agreement in all criteria (κ, 0.85-0.95). Across criteria, a consistent number of patients progressed (23-26) and was associated with significantly poor OS (ps ≤ 0.049). Responders by any criteria showed longer PFS (ps ≤ 0.009), and 1-year OS (ps ≤ 0.01). Choi criteria linked to significantly better OS without landmark (p = 0.003), with 1-year OS rates at 76.9% for responders vs 38.1% for non-responders. Cox analysis identified responders by Choi criteria as a significant OS predictor. Choi criteria identified more early responders than RECIST 1.1 and mRECIST, significantly correlating with improved OS. Choi criteria could be considered as a formal response assessment criterion for the emerging atezo-bev systemic treatment. For atezo-bev treatment of advanced HCC, more comprehensive response criteria, such as Choi criteria, could be effective in identifying early responders and predicting survival outcomes along with RECIST 1.1 and mRECIST. Choi criteria identified a higher rate of early responders compared to mRECIST and RECIST1.1 following atezo-bev treatment. Responders by all criteria had longer PFS and 1-year OS, and only those by Choi criteria experienced longer OS without landmark time. Choi criteria, with RECIST 1.1 and mRECIST, is an effective response assessment tool for atezo-bev treatment.