This study tested whether three different cement layer thicknesses (60, 120 and 180 μm) would provide the same bonding capacity between adhesively luted lithium disilicate and human dentin. Ceramic blocks were cut to 20 blocks with a low-speed diamond saw under cooling water and were then cemented to human flat dentin with an adhesive protocol. The assembly was sectioned into 1 mm2 cross-section beams composed of ceramic/cement/dentin. Cement layer thickness was measured, and three groups were formed. Half of the samples were immediately tested to evaluate the short-term bond strength and the other half were submitted to an aging simulation. The microtensile test was performed in a universal testing machine, and the bond strength (MPa) was calculated. The fractured specimens were examined under stereomicroscopy. Applying the finite element method, the residual stress of polymerization shrinkage according to cement layer thickness was also calculated using first principal stress as analysis criteria. Kruskal–Wallis tests showed that the ‘‘cement layer thickness’’ factor significantly influenced the bond strength results for the aged samples (p = 0.028); however, no statistically significant difference was found between the immediately tested groups (p = 0.569). The higher the cement layer thickness, the higher the residual stress generated at the adhesive interface due to cement polymerization shrinkage. In conclusion, the cement layer thickness does not affect the immediate bond strength in lithium disilicate restorations; however, thinner cement layers are most stable in the short term, showing constant bond strength and lower residual stress.