This article describes the application of response surface methodology to the development of a procedure for mercury determination by cold vapor atomic absorption spectroscopy (CV-AAS) in lipstick samples after digestion by Kieldahl method. A Box-Hunter matrix was used to find optimal conditions for the procedure through a response surface study. Three variables "time, temperature, volume of HNO3 acid" were regarded as factors in the optimization study. In optimized condition, the linearity range was 0.2 ÷ 15 µg/L while the limit of detection (LOD) and limit of quantification (LOQ) were 0.17 and 0.57 ppb, respectively. This method presented good precision (RSD ≤ 8.2%) and good recovery (81%-109%). This method was applied to determine mercury in seven lipstick samples and the highest mercury content was 0.229 ± 0.016 mg/kg.