The purpose of the current study was to test theory-based predictions of mediators and moderators of treatment effects of a pilot randomized controlled trial, which aimed to increase adherence to preventive medication in stroke survivors via addressing both automatic (i.e., habitual responses) and reflective (i.e., beliefs and value systems) aspects of medication-taking behavior. Sixty-two stroke survivors were randomly allocated to either an intervention or control group. Intervention participants received a brief 2-session intervention aimed at increasing adherence via (a) helping patients establish better medication-taking routines using implementation intentions plans (automatic), and (b) eliciting and modifying any mistaken patient beliefs regarding medication and/or stroke (reflective). The control group received similar levels of non-medication-related contact. Primary outcome was adherence to antihypertensive medicine measured objectively over 3 months using an electronic pill bottle. Secondary outcome measures included self-reported adherence (including forgetting) and beliefs about medication. Intervention participants had 10% greater adherence on doses taken on schedule (intervention, 97%; control, 87%; 95% CI [0.2, 16.2], p = .048), as well as significantly greater increases in self-reported adherence and reductions in concerns about medication. Treatment effects were mediated by reductions in both forgetting and concerns about medication, and moderated by the presence of preexisting medication-taking routines. Addressing both automatic and reflective aspects of behavior via helping stroke survivors develop planned regular routines for medication-taking, and addressing any concerns or misconceptions about their medication, can improve adherence and thus potentially patient outcomes.