Gold (Au) in ores typically appears in the form of granular gold or lattice gold and occurs in different states in different types of gold ores. The particle effect caused by its unique malleability challenges the analysis technique for gold samples. The preparation process involves numerous difficult-to-control factors, directly impacting the accuracy and stability of gold analysis. Taking the gold element in ores as the research object, an analytical method for determining gold in ores using the foam plastic adsorption-atomic absorption spectroscopy technique is established. Qualitative judgments and quantitative calculations are made for the key factors in the sample testing process, and an optimization model based on grey relational analysis-response surface methodology is proposed. The important steps in the sample analysis technique, such as roasting, digestion, enrichment, and elution, are discussed. The optimization factors of the roasting method, digestion acidity, enrichment time, and thiourea concentration are determined. An orthogonal experiment is designed, and correlation analysis is carried out. Grey relational coefficients are calculated, and range analysis is used to qualitatively determine the significance of each optimization factor. A significant level table is constructed by combining central composite design with response surface methodology. A predictive model is established using a quadratic polynomial regression equation, and significance analysis is performed. Three-dimensional response surface plots and two-dimensional contour plots are used to fit and analyze the experimental data, determining the optimal parameters of the optimization model as follows: roasting method stepwise segmented roasting, digestion acidity-10.58%, enrichment time-40.00 min, and thiourea concentration-11.65 g·L^-1. Experimental results indicate that under the conditions of the optimized model, when preparing and testing national first-class standard gold ore materials, the method has a detection limit of 0.021, a determination limit of 0.077, and spike recovery rates ranging from 91.6% to 104.5%. The accuracy and precision meet the requirements of GB/T 20899.2—2019 quality control. Furthermore, method validation and comparison are performed on external samples from the Xiqinling area in Shaanxi and the Meichuan area in Gansu. The relative deviation does not exceed 10% for all cases, and the evaluation results are excellent, indicating that the proposed detection optimization model remains accurate and reliable for practical samples, demonstrating correctness and scientific validity. This study presents a new method for the rapid, accurate, and convenient analysis of the Au element in geological and mineral laboratories, providing new ideas for optimizing multi-objective parameter combinations in inspection and testing. It also contributes to accurately testing the new round of strategic mineral exploration.