The analytical method of trace gold in geochemical samples by graphite furnace atomic absorption spectrometry(GFAAS) has been widely used. However, it is hard to figure out the appropriate matching parameters in practice in terms of the configuration of its determination conditions and temperature-rising program parameters. Due to this reason, it is of great practical significance to work out effective parameters quickly and accurately. The single factor test is carried out with lamp current, ashing temperature, and atomization temperature as independent variables and its parameters are set to be 6~8 mA, 300~500 ℃ and 2 200~2 400 ℃ respectively; The Box-Behnken test is done according to the Response Surface Methodology (RSM). The influence of three-factor and three-level surface design on the response value (absorbance) is analyzed; the significance level table is prepared, and the response surface test is completed; a prediction model of the quadratic polynomial regression equation is built analyze the significance. F=43.95, p<0.000 1 means that the model has high significance with its correlation coefficient of 0.985 1.Thefact that the correction coefficient of determination is 0.962 6 indicates that the model can explain more than 95% of the changes in response values; the response surface and contour map are drawn to regress and fit the test data. Judgement and analysis are made according to the shape of the response surface and the steepness of the contour. These optimal parameters as lamp current 7.12 mA, ashing temperature 412.32 ℃, and atomization temperature 2 311.61 ℃ are worked out. The results suggest that under an optimized condition, the average absorbance of national first-class reference material GBW07246a is 0.101 2, basically consistent with the predicted value of 0.108 0, and the relative error remains 6.30%; six national first-class reference materials, including GBW07243b are selected for 12 repeated tests, and the standard curve is drawn, and the results are read. The logarithmic deviation between the average value and the standard value of each reference material is less than 0.05, the relative standard deviation (RSD, n=12) is less than 10%, and both the accuracy and precision comply with GB/T 27417—2017 (Guide for confirmation and verification of chemical analysis for conformity assessment), indicating that the parameters of conditions for the determination of trace gold by graphite furnace atomic absorption spectrometry based on RSM model are accurate and reliable, which proves the correctness and feasibility of this model and achieves a good optimization result. This method is expected to be applied to determine and analyze other elements and in the method research of instrument analysis platform to find the optimal analysis and test conditions.