Background: A previous study of therapist-assisted Internet-delivered cognitive behaviour therapy (ICBT) for generalized anxiety (Paxling et al., 2013) identified eight distinct therapist behaviours in ICBT (task reinforcement, self-efficacy shaping, task prompting, alliance bolstering, psychoeducation, empathetic utterances, deadline flexibility, and self-disclosure). It is unknown how generalizable these behaviours are across ICBT programs. Aims: We systematically examined the frequency of these eight therapist behaviours and additional newly identified behaviours in e-mails sent to patients during the course of ICBT for depressive symptoms. We also conducted exploratory analyses to examine relationships between therapist behaviours, symptom improvement, and therapeutic alliance. Method: Data was obtained from a previously published open trial (Hadjistavropoulos et al., 2014). A total of 1013 e-mails sent from therapists (n = 24) to patients (n = 41) during ICBT for depressive symptoms were analyzed. Therapist behaviours were correlated with symptom change scores and ratings of therapeutic alliance at mid- and post-treatment. Results: Therapist behaviours described by Paxling et al. were reliably identified in the e-mails using qualitative content analysis; the frequencies of these behaviours differed, however, from the Paxling et al. study and three additional therapist behaviours were identified (administrative statements, questionnaire feedback, asking clarifying questions). Several therapist behaviours (e.g. administrative statements, task prompting) were associated with lower symptom improvement at post-treatment. Questionnaire feedback and task reinforcement were associated with higher patient ratings of therapeutic alliance. Conclusions: The study provides partial support for the generalizability of therapist-assistance across ICBT programs. Experimental research is needed to examine the impact of varying therapist-assistance on patient outcomes.