Abstract Background Liver resection is effective for hepatocellular carcinoma (HCC) exceeding the Milan criteria in selected patients. However, the benefit of anatomical resection (AR) versus non-anatomical resection (NAR) has not been clarified in this patient subgroup. This study aimed to compare outcomes between AR and NAR for HCC exceeding the Milan criteria. Methods Data on consecutive patients with HCC exceeding the Milan criteria who underwent liver resection with curative intent over a recent 6-year interval were extracted from a prospective single-centre HCC database and examined retrospectively. The postoperative outcomes of patients were compared before and after propensity score matching. Results Some 546 patients were included: 264 in the AR and 282 in the NAR group. In the original cohort, the AR group contained more patients with larger tumours, multiple tumours, macroscopic portal vein tumour thrombi, incomplete tumour capsules and microscopic vascular invasion. After propensity score matching, 177 pairs of patients were selected. The baseline data, including liver function and tumour burden, were similar in the matched groups. The 3-year recurrence-free survival rate was comparable between the matched NAR and AR groups (36·5 versus 28·5 per cent; P = 0·448). Similar results were observed for 3-year overall survival (57·5 versus 50·3 per cent; P = 0·385), recurrence patterns and early recurrence rates (57·6 per cent versus 59·9 per cent; P = 0·712). Conclusion AR and NAR achieved favourable and similar outcomes for HCC exceeding the Milan criteria in selected patients.