Abstract: In the application of Good Agriculture Practice (GAP) to the vegetable demonstration bases, there is more subjectivity in the aggregation of linguistic evaluation information from the experts. In order to solve the loss and distortion, this research is illustrated by an example relevant to the agricultural product safety evaluation on three export spinach sites that are put on records in Shandong Province. A comprehensive multi-attribute group decision-making approach based on linguistic information processing was proposed, an agricultural product safety evaluation index system was put forward according GAP standard, the statistical evaluation data were gathered from GAP experts, and they are transformed into two-tuple linguistic information representation, then a decision-making model was developed to obtain the values from comprehensive group evaluation, further, sites selection and decision making can be achieved on the basis of above analysis. The above research indicates that the linguistic information can be effectively aggregated with less loss. Therefore, the proposed approach is computationally simple, which will promote the extension and application of GAP, and realize the target of agricultural product safety.