Purpose To prospectively describe magnitude-based multi-echo gradient-echo hepatic proton density fat fraction (PDFF) inter-examination precision at 3 Tesla (T). Materials and Methods In this prospective, Institutional Review Board-approved, Health Insurance Portability and Accountability Act (HIPAA) compliant study, written informed consent was obtained from 29 subjects (body mass indexes>30 kg/m2). Three 3T MRI examinations were obtained over 75-90min. Segmental, lobar, and whole liver PDFF were estimated (using three, four, five, or six echoes) by magnitude-based multi-echo MRI in colocalized regions of interest. For estimate (using three, four, five, or six echoes), at each anatomic level (segmental, lobar, whole liver), three inter-examination precision metrics were computed: intra-class correlation coefficient (ICC), standard deviation (SD), and range. Results Magnitude-based PDFF estimates using each reconstruction method showed excellent inter-examination precision for each segment (ICC≥0.992; SD≤0.66%; range≤1.24%), lobe (ICC≥0.998; SD≤0.34%; range≤0.64%), and the whole liver (ICC=0.999; SD≤0.24%; range≤0.45%). Inter-examination precision was unaffected by whether PDFF was estimated using three, four, five, or six echoes. Conclusion Magnitude-based PDFF estimation shows high inter-examination precision at segmental, lobar, and whole liver anatomic levels, supporting its use in clinical care or clinical trials. The results of this study suggest that longitudinal hepatic PDFF change greater than 1.6% is likely to represent signal rather than noise. J. Magn. Reson. Imaging 2014;39:1265-1271. © 2013 Wiley Periodicals, Inc. Copyright © 2013 Wiley Periodicals, Inc.