This in vitro study assessed the marginal microleakage on class V cavities prepared with aluminum oxide air abrasion and restored with different glass ionomer cements. The cavities were prepared on the buccal and lingual surfaces of 15 sound third molars with an air- abrasion device (Kreativ Mach 4.1; New Image) using a 27.5-microm aluminum oxide particle stream, and were assigned to 3 groups of 10 cavities each. The restorative materials were: group I, a conventional glass ionomer cement (Ketac-Fil); groups II and III, resin-modified glass ionomer cements (Vitremer R and Fuji II LC, respectively). After placement of the restorations, the teeth were stored in distilled water at 37 degrees C for 24 h, polished and then submitted to a thermocycling regimen of 500 cycles, isolated, immersed in 0.2% Rhodamine B solution for 24 h, included and serially sectioned. Microleakage was assessed by viewing the specimens under an optical microscope connected to a color video camera and a computer. The images obtained were digitized and analyzed for microleakage using software that allows for a standard quantitative assessment of dye penetration in millimeters. Statistical analysis was done using the Kruskall-Wallis and Wilcoxon tests. Means of dye penetration (%) were: occlusal - I: 25.76 +/- 34.35, II: 20.00 +/- 42.16, III: 28.25 +/- 41.67; cervical - I: 23.72 +/- 41.84; II: 44.22 +/- 49.69, III: 39.27 +/- 50.74. No statistically significant differences (p>0.05) were observed among either the glass ionomer cements or the margins. In conclusion, class V cavities restored with either conventional or resin-modified glass ionomer cements after preparation with aluminum oxide air abrasion did not show complete sealing at the enamel and dentin/cementum margins.