Despite advances in psychotherapy research showing an evidence-base for psychodynamic psychotherapy (PDT) in adolescents, developmentally specific treatment characteristics are under-researched. We aimed to identify interaction structures (IS: reciprocal patterns of in-session interactions involving therapist interventions, patient behaviors, and the therapeutic relationship) and assess associations between IS and outcome. The study cohort comprised 43 adolescents (Mage = 13.02 years) with nonclinical, internalizing, and comorbid internalizing–externalizing problems in PDT. A total of 123 sessions from different treatment phases were rated based on the Adolescent Psychotherapy Q-Set (APQ). Outcome was assessed with the Brief Problem Monitor-Youth (BPM-Y) administered repeatedly over the treatment course. Principal component analysis of APQ items resulted in five IS, named “Negative Therapeutic Alliance”, “Demanding Patient, Accommodating Therapist”, “Emotionally Distant Resistant Patient”, “Inexpressive Patient, Inviting Therapist”, and “Exploratory Psychodynamic Technique” (EPT). Multilevel modeling analyses with Bayesian Markov chain Monte Carlo (MCMC) estimations indicated a two-way interaction effect between EPT and problem levels at baseline such that patients with lower problems at baseline showed good outcome in the context of EPT, whereas an inverse relationship was found for patients with higher problems. Findings provide empirical evidence for characteristic components of PDT for adolescents and preliminary answers about who benefits from psychodynamic techniques.