Aim The aim of this clinical trial was to evaluate and compare the performance of a universal adhesive with different adhesive strategies in the restoration of non-carious cervical lesions (NCCLs) over a 36-month period. Material and method One hundred sixty-five NCCLs in 35 patients (13 female, 22 male) with at least 3 lesions each were included in this study. Three groups were formed according to the adhesive strategy used (n = 55): selective-etch mode, etch-and-rinse mode, or self-etch mode of a universal adhesive, Single Bond Universal. The same nanofilled resin composite, Filtek Ultimate, was used for all restorations by a single operator. The restorations were evaluated by two calibrated examiners at baseline and at 6, 12, 18, 24, and 36 months in accordance with the modified USPHS criteria. The chi-square test was used for intergroup comparison and Cochran's Q test for intragroup comparison (proportional to = 5%). Results At 36 months, the recall rate was 98.1% and three restorations, one from each group, had failed because of retention loss (P > 0.05). The self-etch mode group showed 17 bravo scores for both marginal staining and marginal adaptation after 3 years, which was significantly different from the selective-etch and etch-and-rinse groups (P < 0.05). For all groups, the only statistically significant difference was found when baseline and 36-month evaluations were compared in terms of marginal staining (P = 0.000). When the marginal adaptation values at 36 months were compared with those at the baseline, statistically significant differences were found in the etch-and-rinse and self-etch mode groups (P < 0.05). Neither secondary caries nor postoperative sensitivity was observed at any recall. Conclusion All adhesive modes showed similar retention rates. Although all restorations were clinically acceptable, restorations in self-etch mode showed less satisfying performance for marginal staining and marginal adaptation.