In a randomized controlled trial, we found that suicidal patients who received Collaborative Assessment and Management of Suicidality (CAMS) reported greater improvements in suicidal ideation and mental health distress compared to participants who received treatment as usual (TAU). Here, we explored moderators and mediators of the effectiveness of CAMS. Compared to TAU, CAMS was more effective in reducing suicidal ideation when the working alliance, in particular its bond subcomponent, was low. In terms of reducing mental health distress, CAMS was superior to TAU only for participants who did not use illicit drugs and, more tentatively, only for patients without borderline personality traits. We suggest that CAMS may repair a difficult vantage point in terms of poor working alliance in patients with suicide ideation. To obtain superior benefits of CAMS upon more general mental health distress in patients with drug abuse or borderline traits, these problems may need to be more explicitly targeted in parallel.