Background: Internet-based mental health interventions are considered effective in providing low-threshold support for people with mental health disorders. However, there is a lack of research investigating the transferability of such online programs into routine care settings. Low treatment adherence and problems with technical implementation often limit a successful transfer into clinical routines. This naturalistic study aims to identify influencing factors on program adherence in patients who participated in an online intervention during inpatient or day-clinic psychotherapeutic treatment. Methods and findings: In a naturalistic study, we investigated the transferability of the transdiagnostic psychodynamic online self-help program KEN-Online, which includes eight consecutive units. Between May 2017 and October 2018, patients who received inpatient or day-clinic psychotherapeutic treatment at the Department of Psychosomatic Medicine and Psychotherapy in the University Medical Center Mainz have been offered to use KEN-Online. Of the n = 749 patients who were admitted to the clinic, n = 239 patients (32%) registered for participation in KEN-Online. While 46.9% of the participants did not complete any unit (inactive participants), 53.1% completed at least the first unit (active participants). Age, number of diagnoses, and symptom severity were associated with (in)active participation. Adherence decreased over time resulting in only 17 participants (7.6%) who completed all units. None of the sociodemographic and medical characteristics proved to be significant predictors of adherence. Analyses of effectiveness showed a significant reduction of anxiety and depression in active participants in the course of participation, with higher improvements in participants that completed more than half of the units. Conclusion: Adherence to the online self-help program KEN-Online was lower in the naturalistic setting than in a previous clinical trial, but was still associated with greater program effectiveness. Adherence-promoting measures are crucial to increase the effectiveness of such interventions in clinical settings.