Purpose: Using an optical sensor, to automatically evaluate the marginal seal of restorations placed with 21 adhesive systems of all four adhesive categories in cylindrical cavities of bovine dentin applying different outcome variables, and to evaluate their discriminatory power. Materials and Methods: Twenty-one adhesive systems were evaluated: three 3-step etch-and-rinse systems, three 2-step etch-and-rinse systems, five 2-step self-etching systems, and ten 1-step self-etching systems. All adhesives were applied in cylindrical cavities in bovine dentin together with Tetric Ceram (n=8). In the control group, no adhesive system was used. After 24 h of storage in water at 37°C, the surface was polished with 4000-grit SiC paper, and epoxy resin replicas were produced. An optical sensor (FRT MicroProf) created 100 profiles of the restoration margin, and an algorithm detected gaps and calculated their depths and widths. The following evaluation criteria were used: percentage of specimens without gaps, the percentage of gap-free profiles in relation to all profiles per specimen, mean gap width, mean gap depth, largest gap, modified marginal integrity index MI. The statistical analysis was carried out on log-transformed data for all variables with ANOVA and post-hoc Tukey's test for multiple comparisons. The correlation between the variables was tested with regression analysis, and the pooled data according to the four adhesive categories were compared by applying the Mann-Whitney nonparametric test (p 0.05). Results: For all the variables that characterized the marginal adaptation, there was a great variation from material to material. In general, the etch-and-rinse adhesive systems demonstrated the best marginal adaptation, followed by the 2-step self-etching and the 1-step self-etching adhesives; the latter showed the highest variability in test results between materials and within the same material. The only exception to this rule was Xeno IV, which showed a marginal adaptation that was comparable to that of the best 3-step etch-and-rinse systems. Except for the variables "largest gap" and "mean gap depth", all the other variables had a similar ability to discriminate between materials. Pooled data according to the four adhesive categories revealed statistically significant differences between the one-step self-etching systems and the other three systems as well as between two-step self-etching and three-step etch-and-rinse systems. Conclusions:With one exception, the one-step self-etching systems yielded the poorest marginal adaptation results and the highest variability between materials and within the same material. Except for the variable "largest gap", the percentage of continuous margin, mean gap width, mean gap depth, and the marginal integrity index MI were closely related to one another and showed - with the exception of "mean gap depth" - similar discriminatory power. Palabras clave: automated marginal analysis, optical sensor, adhesive systems, gap width, gap depth, MI index