Background Children with hepatoblastoma routinely undergo repetitive surveillance imaging, with CT scans for several years after therapy, increasing the risk of radiation-induced cancer. Objective The purpose of this study was to determine the utility of surveillance CT scans compared to serum alpha-fetoprotein (AFP) levels for the detection of hepatoblastoma relapse. Materials and methods This was a retrospective study of all children diagnosed with AFP-positive hepatoblastoma from 2001 to 2011 at a single institution. Results Twenty-six children with hepatoblastoma were identified, with a mean age at diagnosis of 2 years 4 months (range 3 months to 11 years). Mean AFP level at diagnosis was 132,732 ng/ml (range 172.8–572,613 ng/ml). Five of the 26 children had hepatoblastoma relapse. A total of 105 imaging exams were performed following completion of therapy; 88 (84%) CT, 8 (8%) MRI, 5 (5%) US and 4 (4%) FDG PET/CT exams. A total of 288 alpha-fetoprotein levels were drawn, with a mean of 11 per child. The AFP level was elevated in all recurrences and no relapses were detected by imaging before AFP elevation. Two false-positive AFP levels and 15 false-positive imaging exams were detected. AFP elevation was found to be significantly more specific than PET/CT and CT imaging at detecting relapse. Conclusion We recommend using serial serum AFP levels as the preferred method of surveillance in children with AFP-positive hepatoblastoma, reserving imaging for the early postoperative period, for children at high risk of relapse, and for determination of the anatomical site of clinically suspected recurrence. Given the small size of this preliminary study, validation in a larger patient population is warranted.