Objective This study examines relationships among different aspects of therapeutic alliance with treatment outcome, adherence and attrition in internet delivered cognitive behavioral therapy (ICBT) for panic disorder. Methods: We examined alliance-outcome relationships in ICBT (N = 74) using a newly developed self-report alliance measure that disentangles alliance with program content (Internet Patients Experience of Attunement and Responsiveness with the program; I-PEARp) and with the therapist (I-PEARt). We compared ICBT outcomes of patient rated and therapist-rated alliance with conventional alliance scales (WAI-6 and WAI-T). Results: Consistent with our hypothesis, I-PEARp and I-PEARt distinguished between different aspects of the alliance and predicted outcomes better than standard alliance scales. Furthermore, higher ratings of I-PEARp were associated with subsequent lower symptoms and lower symptoms were associated with higher subsequent alliance. In contrast, I-PEARt predicted adherence, but not symptoms. Although therapists ratings of alliance (thI-PEAR) improved significantly during treatment, they did not predict subsequent symptoms, adherence, or dropout. Conclusion: Results indicate that the patient experience of the alliance in ICBT includes two aspects, each of which uniquely contributes to outcomes; patient connection to the program is related to symptom outcomes whereas the dyadic relationship with the therapist serves as the glue to allow the treatment to hold.