(1) Aim of this in-vitro study was to investigate the handling of proximal-cervical undermined enamel margins on adhesive performance of differently fabricated and differently cemented ceramic inlays and partial crowns (2) Methods: 192 extracted third molars received mod (n=96) and partial crown (n=96) preparations. A mesial 2x2x4 mm cervical groove was created in dentin to simulate a deeper (dentin) caries excavation. This dentin groove was either left (G/groove), filled with composite (F/filling) or completely removed (D/dentin). Labside (e.max Press) and chairside (e.max CAD) inlays and partial crowns were adhesively luted with Syntac/Variolink Esthetic (SV) or Adhese Universal/Variolink Esthetic (AV). Initially, and again after thermomechanical loading (TMB: 1 million cycles at 50 N, 25,000 thermocycles at 5°C/55°C), epoxy replicas were examined for marginal gaps using scanning electron microscopy (200x magnification). Light microscopy (10x magnification) was used to evaluate proximal cervical crack propagation. (3) Results: Regardless of the adhesive system, D groups generally showed significantly lower marginal quality (p<0.05), with the universal adhesive performing better than the multi-step adhesive system (p<0.05). Subgroups G and F were similar in marginal quality (p>0.05) and not worse than the controls (p>0.05) regardless of the adhesive system, but showed less cracking in F than in G (p<0.05). In general, fewer cracks were observed in CAD/CAM restorations than in laboratory-fabricated restorations (p<0.05). Partial crowns showed better marginal quality and less cracking than inlays (p<0.05). (4) Conclusions: If the dentin level is lower than the enamel level in ceramic preparations after caries excavation in the proximal box, the resulting undermined enamel should not be removed.