Abstract Purpose: Atezolizumab + bevacizumab is the new standard of care for systemic treatment-naïve, unresectable hepatocellular carcinoma (HCC). This exploratory study investigated on-treatment alpha-fetoprotein (AFP) response as a potential surrogate biomarker of prognosis for the combination therapy. Experimental Design: Data from Group A of the phase Ib GO30140 study were used to identify the optimal time for AFP measurement and AFP cutoffs to differentiate patients by their best confirmed response per independent review facility–assessed RECIST (IRF-RECIST) version 1.1: responders from nonresponders and patients with disease control from primary progressors. We applied these cutoffs to independent data from the atezolizumab + bevacizumab arm of the phase III IMbrave150 trial to distinguish patients based on (i) overall survival (OS) and progression-free survival (PFS) per IRF-RECIST 1.1 and (ii) best confirmed response per IRF-RECIST 1.1. Results: We derived AFP cutoffs of ≥75% decrease and ≤10% increase from baseline at 6 weeks to identify responders and those who had disease control, respectively. These cutoffs had high sensitivity and specificity in GO30140. In IMbrave150 patients, sensitivity was 0.59 and specificity was 0.86 for the ≥75% decrease AFP cutoff; the sensitivity was 0.77 and specificity was 0.44 for the ≤10% increase AFP cutoff. Both AFP cutoffs were associated with longer OS and PFS, particularly in patients with hepatitis B virus etiology (HR < 0.5; P < 0.01). Conclusions: AFP response at 6 weeks after initiating treatment is a potential surrogate biomarker of prognosis for patients with HCC receiving atezolizumab + bevacizumab. See related commentary by Cappuyns and Llovet, p. 3405