Recommendation Agents (RAs ) are″virtual consultants″provided by on-line shops . RAs encourage consumers to make purchase decisions quickly . In order to determine how RA use influences consumer on-line decision making , this paper develops a conceptual model describing the relationship among three types of variables : RA characteristics , RA evaluation and RA power . The model is tested based on a large sample data from Internet consumers in China . The results show that : ( 1 ) RA use influences the processes and outcomes of consumer decision making and the extent of this influence ( RA power) is determined by RA evaluation , i .e . , the consumer's perceived value of the RAs ( RA perceived value) and their trust in RAs (RA trust);(2) RA perceived value is determined by the general public perception of the website in which the RAs are embedded , the customization ( the most important factor ) and presentation of recommendations;and (3) RA trust is determined by RA perceived value , the general public perception of the website , the customization of recommendations and consumer's familiarity with RAs . Of these , RA perceived value is the most important factor .