Purpose: This study compared the marginal gaps of computer-aided design/computeraided manufacture (CAD/CAM)-fabricated all-ceramic crowns constructed from scanned impressions and models and with two different occlusal reduction designs. Materials and Methods: Two typodont mandibular first molars were prepared to receive CAD/CAM- fabricated all-ceramic crowns. Both molars were prepared to ideal crown reduction, the first with anatomical occlusal reduction (AOR) and the second with completely flat occlusal reduction (FOR). Nine polyvinyl siloxane impressions (PVS) were taken, and nine stone replicas fabricated for each preparation. All impressions and stone models were scanned using a laser scanner (Planmeca Planscan, E4D technologies), and 36 lithium disilicate (IPS e.max CAD) crowns were milled. The marginal gap was measured in four locations using a light stereomicroscope. Results: Crowns constructed from preparations with both occlusal reduction designs demonstrated similar marginal gaps (FOR = 97.98; AOR = 89.12; P = .739). However, all crowns constructed from scanned impressions presented significantly larger marginal gaps than the crowns fabricated from scanned models (impressions = 109.26; models = 77.84; P = .002). Conclusion: Scanning stone models produced all-ceramic crowns with significantly smaller marginal gaps than scanning impressions, irrespective of the occlusal reduction design.