This study aimed to address the time, cost, and ethical issues associated with traditional animal experiment-based observational methods by utilizing in silico Physiologically Based Pharmacokinetic modeling to predict veterinary drug residues in livestock products and validate them against observational data. Using PK-sim software, we modeled the physiological conditions of pigs to predict the depletion of ceftiofur and spiramycin. We evaluated the ceftiofur (3 mg, 6 mg) and spiramycin models by comparing them with observational data using residuals, MSE, and R-squared values. Specifically, the R-squared values for the ceftiofur models were all negative, indicating poor predictive power. For Ceftiofur (3 mg), the R-squared value was <0 with MSE of 611.3764, and for Ceftiofur (6 mg), it was <0 with MSE of 2447.982, highlighting significant discrepancies. Similar shortcomings were observed in the spiramycin models, with an R-squared value of <0 . These discrepancies can be attributed to inaccuracies in literature data, limited physicochemical data, inadequate consideration of inter-individual differences, mismatches between experimental and model conditions, and limitations of benchmark observational experiments. This underscores the critical importance of enhancing data quality and refining modeling approaches. Future research should focus on validating in silico techniques across diverse animal models and drugs to broaden their applicability in safety assessments. Ultimately, leveraging in silico techniques is crucial for establishing a scientifically robust safety management system for livestock products, overcoming the constraints of current observational experimental methods.