Background Mindfulness-based cognitive therapy (MBCT) is an evidence-based group-based psychological treatment in oncology, resulting in reduction of depressive and anxiety symptoms. Internet-based MBCT (eMBCT) has been found to be an effective alternative for MBCT. The therapeutic alliance (the bond between therapist and patient,) is known to have a significant impact on psychological treatment outcomes, including MBCT. A primary concern in the practice of eMBCT is whether a good therapeutic alliance can develop. Although evidence for the beneficial effect of therapist assistance on treatment outcome in internet-based interventions (IBIs) is accumulating, it is still unclear whether the therapeutic alliance is related to outcome in IBIs. Objective This study aimed to (1) explore whether early therapeutic alliance predicts treatment dropout in MBCT or eMBCT, (2) compare the development of the therapeutic alliance during eMBCT and MBCT, and (3) examine whether early therapeutic alliance is a predictor of the reduction of psychological distress and the increase of mental well-being at posttreatment in both conditions. Methods This study was part of a multicenter randomized controlled trial (n=245) on the effectiveness of MBCT or eMBCT for distressed cancer patients. The therapeutic alliance was measured at the start of week 2 (ie, early therapeutic alliance), week 5, and week 9. Outcome measures were psychological distress, measured with the Hospital Anxiety and Depression Scale, and mental well-being, measured with the Mental Health Continuum-Short Form. Results The strength of early therapeutic alliance did not predict treatment dropout in MBCT or eMBCT (B=−.39; P=.21). Therapeutic alliance increased over time in both conditions (F2,90=16.46; Wilks λ=0.732; P<.001). This increase did not differ between eMBCT and MBCT (F1,91=0.114; P=.74). Therapeutic alliance at week 2 predicted a decrease in psychological distress (B=−.12; t114=−2.656; P=.01) and an increase in mental well-being (B=.23; t113=2.651; P=.01) at posttreatment. The relationship with reduction of psychological distress differed between treatments: a weaker early therapeutic alliance predicted higher psychological distress at posttreatment in MBCT but not in eMBCT (B=.22; t113=2.261; P=.03). Conclusions A therapeutic alliance can develop in both eMBCT and MBCT. Findings revealed that the strength of early alliance did not predict treatment dropout. Furthermore, the level of therapeutic alliance predicted reduced psychological distress and increased mental well-being at posttreatment in both conditions. Interestingly, the strength of therapeutic alliance appeared to be more related to treatment outcome in group-based MBCT than in eMBCT. Trial Registration ClinicalTrials.gov NCT02138513; https://clinicaltrials.gov/ct2/show/NCT02138513