Background: Internet-based cognitive behavioral therapy (ICBT) is a form of guided self-help that has been found to be effective for addressing several problems. The target for this type of therapy is usually restricted to one specific disorder. Tailoring the treatment widens the scope of ICBT in that it can address comorbid conditions directly. Objectives: The working, or therapeutic, alliance has been found to predict outcome in studies of face-to-face therapy. The extent to which these findings apply to ICBT is largely unknown. We therefore decided to find out whether the working alliance could predict outcome in tailored ICBT for anxiety disorders. Methods: Data were obtained from the treatment group (n=27) in a randomized controlled trial aiming to test the effects of tailored ICBT for anxiety disorders. The forthcoming study was designed to test the hypothesis that the working alliance measured both pre-treatment and early in treatment (week 3) can predict treatment outcome as measured by the Clinical Outcomes in Routine Evaluation–Outcome Measure (CORE-OM) in a heterogeneous group of patients with anxiety disorders (n=27). Results: Working alliance measured at week 3 into the treatment correlated significantly with the residual gain scores on the primary outcome measure (r=-.47, P=.019, n=25), while expected working alliance pre-treatment did not (r=-.17, P=.42, n=27). Conclusions: These results raise questions about the importance of working alliance in ICBT treatments, and suggest that the working alliance could be important in ICBT.