OBJECTIVE This study was designed to adapt a previously developed in vitro microcosm biofilm model to create carries-affected dentin (CAD) and establish conditions for using the model in bonding studies. METHODS Biofilms were originated from human saliva and grown on dentin discs for 0 (sound dentin), 3, 5, 7, 14, or 21 days under intermittent cariogenic condition (n = 10). At each time point, composite cylinders were bonded to the dentin using self-etch adhesive (Clearfil SE Bond). The response variables were integrated mineral loss (ΔS), lesion depth (LD), shear bond strength (SBS), and failure mode. Data were statistically analyzed (α = 0.05). Bonded interfaces were analyzed by scanning electron microscopy (SEM), and dentin surfaces characterized by infrared spectroscopy (Fourier transform infrared spectroscopy, FTIR). RESULTS Lower ΔS was found for sound dentin than for CAD in all experimental groups, except for the group under cariogenic challenge for 3 days. The SBS to CAD was significantly lower than control for all cariogenic challenge times. Adhesive failures were predominant in all groups. ΔS and LD had a significant negative correlation with SBS. A significant exponential decay in SBS was associated with increased ΔS values. CAD had lower mineral and amide I content and an irregular hybridization interface compared to sound dentin. CONCLUSIONS The microcosm biofilm model was able to artificially induce CAD, which imposed challenge to the bonding of the polymeric adhesive material. CONCLUSIONS Presence of CAD might interfere with the bonding of polymeric materials. The microcosm biofilm model proposed could be useful for preclinical dentin bonding studies.