The way dentin is prepared is known to reduce the bonding efficiency of especially 'mild' self-etch adhesives. Objectives: To investigate to what extent the way of dentin preparation affects the micro-tensile bond strength (µTBS) to dentin ‘immediately' and after aging using medium-term thermo-cycling. Methods: 60 mid-coronal sound dentin specimens cut from extracted human molars were prepared using either a 100-µm diamond bur (bur-cut), 600-grit silicon carbide (SiC) paper, or were fractured, thus not having produced a smear layer. Resin composite (Herculite XRV Ultra, Kerr) was bonded to the surfaces using either the 2-step self-etch adhesives, OptiBond XTR (Kerr) or Clearfil SE Bond (Kuraray), or the 3-step etch-and-rinse adhesive OptiBond FL (Kerr) that served as gold-standard control. After 1-week water-storage at 37°C, specimens were serially cut into 1-mm² stick-shaped micro-specimens, of which per tooth half of them were further subjected to 20,000 thermo-cycles. The µTBS was determined at a crosshead speed of 1.0 mm/min. Failure analysis was performed using stereo-microscopy and SEM. Data were statistically analyzed by ANOVA and post hoc Tukey test. Results: No significant difference in µTBS was found for all three adhesives when bonded to the three differently prepared dentin surfaces (Table; p>0.05). However, when the specimens were aged using thermo-cycling, the two self-etch adhesives, OptiBond XTR and Clearfil SE Bond, revealed a significantly lower µTBS (p<0.05) to bur-cut dentin then to SiC-ground and fractured dentin, at which respectively a thinner and less compact smear layer, or no smear layer was produced. Conclusion: The way dentin is prepared and thus the kind of the smear layer is produced affect the ‘aged' bond strength to dentin.