Background: Evidence of benefit for telehealth for chronic conditions is mixed. Two linked randomized controlled trials tested the Healthlines Service for 2 chronic conditions: depression and high risk of cardiovascular disease (CVD). This new telehealth service consisted of regular telephone calls from nonclinical, trained health advisers who followed standardized scripts generated by interactive software. Advisors facilitated self-management by supporting participants to use Web-based resources and helped to optimize medication, improve treatment adherence, and encourage healthier lifestyles. Participants were recruited from primary care. The trials identified moderate (for depression) or partial (for CVD risk) effectiveness of the Healthlines Service. Objective: An embedded qualitative study was undertaken to help explain the results of the 2 trials by exploring mechanisms of action, context, and implementation of the intervention. Methods: Qualitative interview study of 21 staff providing usual health care or involved in the intervention and 24 patients receiving the intervention. Results: Interviewees described improved outcomes in some patients, which they attributed to the intervention, describing how components of the model on which the intervention was based helped to achieve benefits. Implementation of the intervention occurred largely as planned. However, contextual issues in patients’ lives and some problems with implementation may have reduced the size of effect of the intervention. For depression, patients’ lives and preferences affected engagement with the intervention: these largely working-age patients had busy and complex lives, which affected their ability to engage, and some patients preferred a therapist-based approach to the cognitive behavioral therapy on offer. For CVD risk, patients’ motivations adversely affected the intervention whereby some patients joined the trial for general health improvement or from altruism, rather than motivation to make lifestyle changes to address their specific risk factors. Implementation was not optimal in the early part of the CVD risk trial owing to technical difficulties and the need to adapt the intervention for use in practice. For both conditions, enthusiastic and motivated staff offering continuity of intervention delivery tailored to individual patients’ needs were identified as important for patient engagement with telehealth; this was not delivered consistently, particularly in the early stages of the trials. Finally, there was a lack of active engagement from primary care. Conclusions: The conceptual model was supported and could be used to develop further telehealth interventions for chronic conditions. It may be possible to increase the effectiveness of this, and similar interventions, by attending to the human as well as the technical aspects of telehealth: offering it to patients actively wanting the intervention, ensuring continuity of delivery by enthusiastic and motivated staff, and encouraging active engagement from primary care staff.