The purpose of this study was to investigate relationships between clients’ perception of session impact, working alliance, and counseling satisfaction in the initial counseling stage. All of the clients in this study had a preoccupied attachment style. The data were collected through questionnaires, namely, the Session Evaluation Questionnaire Form 5 (SEQ-5) (Stiles, Gordon, & Lani, 2002), Working Alliance Inventory (WAI) (Horvath & Greenberg, 1989), and Counseling Satisfaction Inventory-Short Form (CSI-S) (Chen & Jian, 2011). A total of 87 clients from 13 Taiwanese universities were recruited through heterogeneous convenience sampling between October 2014 and July 2016. After their third individual counseling session, clients completed the SEQ-5, WAI, and CSI-S. In addition, partial least squares structural equation modeling (PLS-SEM) was used to test the hypothesized path model. The results of this study were as follows: 1) clients’ session impact predicted working alliance (effect size: .84, p < .001); 2) clients’ working alliance predicted counseling satisfaction (effect size: .56, p < .001); 3) clients’ session impact predicted working alliance (effect size: .84, p < .001); 4) clients’ working alliance predicted counseling satisfaction (effect size: .56, p < .001); 5) clients’ session impact indirectly predicted counseling satisfaction through working alliance (effect size: .47, p < .05); and 6) clients’ session impact directly predicted counseling satisfaction (effect size: .31; p < .01). These findings suggest that clients’ working alliance is likely to mediate the relationship between session impact and counseling satisfaction.