A procedure for the determination of As in diesel, gasoline and naphtha at microg L(-1) levels by GFAAS is proposed. Sample stabilization was achieved by the formation of three component solutions prepared by mixing appropriate volumes of the samples propan-1-ol and nitric acid aqueous solution. This mixture resulted in a one-phase medium, which was indefinitely stable. No changes in the analyte signals were observed over several days in spiked samples, proving long-term stabilization ability. The use of conventional (Pd) and permanent (Ir) modification was investigated and the former was preferred. Central composite design multivariate optimization defined the optimum microemulsion composition as well as the temperature program. In this way, calibration using aqueous analytical solutions was possible, since the same sensitivity was observed in the investigated microemulsion media and in 0.2% v/v HNO(3). Coefficients of correlation larger than 0.999 and an As characteristic mass of 22 pg were observed. Recoveries (n=4) obtained from spiked samples were 98+/-4, 99+/-3 and 103+/-5%, and the limits of detection in the original samples were 1.8, 1.2 and 1.5 microg L(-1) for diesel, gasoline and naphtha, respectively. Validation was performed by the analysis of a set of commercial samples by independent comparative procedures. No significant difference (Student's t-test, p<0.05) was observed between comparative and proposed procedure results. The total determination cycle lasted 4 min for diesel and 3 min for gasoline and naphtha, equivalent to a sample throughput of 7 h(-1) for diesel and 10 h(-1) for gasoline and naphtha.