ObjectiveTo investigate the effect of three different impression methods on the marginal fit of all-ceramic crowns. The three methods include scanning silicone rubber impression, cast models, and direct optical impression.MethodsThe polymethyl methacrylate (PMMA) material of a mandibular first molar in standard model was prepared with 16 models duplicated. The all-ceramic crowns were prepared using three different impression methods. Accurate impressions were made using silicone rubber, and the cast models were obtained. The PMMA models, silicone rubber impressions, and cast models were scanned, and digital models of three groups were obtained to produce 48 zirconia all-ceramic crowns with computer aided design/computer aided manufacture. The marginal fit of these groups was measured by silicone rubber gap impression. Statistical analysis was performed with SPSS 17.0 software.ResultsThe marginal fit of direct optical impression groups, silicone rubber impression groups, cast model groups was (69.18±9.47), (81.04±10.88), (84.42±9.96) μm. A significant difference was observed in the marginal fit of the direct optical impression groups and the other groups (P0.05).ConclusionAll marginal measurement sites are clinically acceptable by the three different impression scanning methods. The silicone rubber impression scanning method can be used for all-ceramic restorations.