Background: Non-invasive tests are widely used to diagnose fibrosis in patients with non-alcoholic fatty liver disease (NAFLD), however, the optimal method remains unclear. We compared the accuracy of simple serum models, a serum model incorporating direct measures of fibrogenesis (Hepascore), and Fibroscan®, for detecting fibrosis in NAFLD. Methods: NAFLD patients undergoing liver biopsy were evaluated with Hepascore, NAFLD Fibrosis Score (NFS), FIB-4 and AST-platelet ratio index (APRI), with a subset (n = 131) undergoing Fibroscan®. Fibrosis on liver biopsy was categorized as advanced (F3–4) or cirrhosis (F4). Accuracy was determined by area under receiving operating characteristic curves (AUC). Indeterminate ranges were calculated using published cut-offs. Results: In 271 NAFLD patients, 83 (31%) had F3–4 and 47 (17%) cirrhosis. 6/131 (4%) had an unreliable Fibroscan®. For the detection of advanced fibrosis, the accuracy of Hepascore (AUC 0.88) was higher than FIB-4 (0.73), NFS (0.72) and APRI (0.69) (p < 0.001 for all). Hepascore had similar accuracy to Fibroscan® (0.80) overall, but higher accuracy in obese individuals (0.91 vs 0.80, p = 0.001). Hepascore more accurately identified patients with cirrhosis than APRI (AUC 0.85 vs 0.71, p = 0.01) and NFS (AUC 0.73, p = 0.01) but performed similar to FIB-4 and Fibroscan®. For the determination of F3-4, the proportion of patients in indeterminate area was lower for Hepascore (4.8%), compared to FIB-4 (42%), NFS (36%) and APRI (44%) (p < 0.001 for all). Conclusions: Hepascore has greater accuracy and a lower indeterminate range than simple serum fibrosis tests for advanced fibrosis in NAFLD, and greater accuracy than Fibroscan® in obese individuals.