Internet-based interventions can be effective treatments for anxiety and depression. Meta-analytic evidence suggests that they should be delivered with human support to reach optimal effects. These findings have not consistently been replicated in direct comparisons of supported and unsupported interventions, however. This study examined the role of support in Internet-based problem solving treatment (PST) for symptoms of anxiety and/or depression. Adults with mild to moderate symptoms of anxiety and/or depression were recruited from the general population and randomized to: (1) PST without support (n = 107), (2) PST with support on request (n = 108), (3) PST with weekly support (n = 106), (4) no Internet-based intervention but non-specific chat or email (n = 110), or (5) waitlist control (WLC; n = 106). Primary outcomes were symptoms of anxiety (HADS) and depression (CES-D) measured at baseline and 6 weeks later. Analyses were first based on the intention-to-treat principle (ITT) and repeated with intervention completers. Only participants who received PST with weekly support improved significantly more than WLC for depressive symptoms. Results for anxiety were less robust but in favor of the weekly support condition. The results underscore the importance of structural support in Internet-based interventions for depression and anxiety.