We used a longitudinal actor–partner interdependence model to examine the codeveloping alliance in alliance empowerment therapy (AET; Escudero, 2013), a manualized team-based approach developed in Spain specifically for child welfare-involved youth. In this first evaluation of AET, we sampled 102 adolescents, 83% of whom had been removed from their homes due to abuse or neglect, and 40% of whom were in crisis at the time of referral. Before each session, clients rated their improvement-so-far; after each session, both clients and therapists completed a brief alliance measure, an adaptation of the System for Observing Family Therapy Alliances (SOFTA-s; Friedlander et al., 2006) for individual therapy. In terms of outcome, clients’ improvement ratings were significantly associated with posttreatment changes in overall functioning and personal goal attainment. With respect to the change process, growth was observed in both clients’ and therapists’ ratings of the alliance over 12 sessions, and an increased similarity in alliance ratings was due to more rapid growth in the therapists’ ratings than the clients’ ratings. Dynamic structural equation modeling indicated that at higher levels of adolescent goal attainment, a stronger association was observed between increased therapist-rated alliance and goal attainment. In other words, in the most effective cases, therapists were more responsive to how the adolescents seemed to have experienced the alliance in the previous session. The cocreated alliance perceptions were due to therapist (rather than client) responsiveness, as well as to unspecified aspects of sharing a therapy environment over time, such as familiarity with the process, regular meetings, and so on.