A methodology was developed for the determination of Se and As in samples of crude oil using inductively coupled plasma mass spectrometry with a dynamic reaction cell. The proposed technique of sample preparation showed the advantage of direct dilution in organic solvent, xylene, eliminating previous treatments, such as digestion of samples, and resulting in an easy and fast analytical method. The cell reaction was optimized using methane as the reaction gas. An univariate study was performed with the operating parameters of the cell, the reaction gas flow rate and rejection parameter q, in order to reduce the interferences on isotopes 80Se and 75As. Organic standards for arsenic and selenium dissolved in xylene were used for this study and a 100% reduction efficiency was observed for 80Se. In the case of 75As, it was observed that the use of the cell was not necessary, since the presence of chlorine was not detected in the organic solutions. With the appropriate experimental conditions for the cell, the experimental conditions were optimized by an univariate study. The conditions used were 1350 W (RF power), 0.40 L min(-1) (Ar-nebulizer flow rate) and 0.7 L min(-1) (Ar-auxiliary flow rate). The standard addition technique was employed for calibration and the instrumental limits of detection (LOD) were 0.23 (80Se) and 0.018 (75As) mg kg(-1). In order to validate the methodology, the residual fuel oil certified reference material (CRM) NIST 1634c was analyzed and recoveries of 97% and 94% were obtained for Se and As, respectively. The confidence intervals showed that the results were within the CRM´s range for both elements. The methodology was applied to the analysis of nine samples of Brazilian crude oil. A hydride generation methodology was used for comparison with the proposed methodology. The samples were decomposed in a heated block and pre-reduced with HCl 6 mol L(-1). External calibration curves were used to determine 78Se, 82Se e 75As, resulting in LODs of approximately 0.04 ug L(-1) to determine total Se and 0.0022 ug L (-1) for the As. The accuracy was evaluated also with NIST 1634c, obtaining recoveries of 99.7% (78Se); 99.2% (82Se) e 100.4% (75As). The Snedecor statistical test showed that the results for selenium in all nine oil samples were in agreement by the two techniques at a 95% confidence level. However, only four oils had concordant results for arsenic by both techniques, the other oils showed lower concentrations for the hydride generation methodology in relation to concentrations found by the direct introduction technique. These results were attributed to losses during the digestion and evaporation steps, and this observation showed the importance of the purposed methodology for accurate assessments of total concentrations of Se and As in crude oil and refined products.