Aim: To investigate the effectiveness of self-directed arm interventions in adult stroke survivors. Methods: A systematic review of Medline, EMBASE, CINAHL, SCOPUS and IEEE Xplore up to February 2018 was carried out. Studies of stroke arm interventions were included where more than 50% of the time spent in therapy was initiated and carried out by the participant. Quality of the evidence was assessed using the Cochrane risk of bias tool. Results: A total of 40 studies (n = 1172 participants) were included (19 randomized controlled trials (RCTs) and 21 before–after studies). Studies were grouped according to no technology or the main additional technology used (no technology n = 5; interactive gaming n = 6; electrical stimulation n = 11; constraint-induced movement therapy n = 6; robotic and dynamic orthotic devices n = 8; mirror therapy n = 1; telerehabilitation n = 2; wearable devices n = 1). A beneficial effect on arm function was found for self-directed interventions using constraint-induced movement therapy (n = 105; standardized mean difference (SMD) 0.39, 95% confidence interval (CI) −0.00 to 0.78) and electrical stimulation (n = 94; SMD 0.50, 95% CI 0.08–0.91). Constraint-induced movement therapy and therapy programmes without technology improved independence in activities of daily living. Sensitivity analysis demonstrated arm function benefit for patients >12 months poststroke (n = 145; SMD 0.52, 95% CI 0.21–0.82) but not at 0–3, 3–6 or 6–12 months. Conclusion: Self-directed interventions can enhance arm recovery after stroke but the effect varies according to the approach used and timing. There were benefits identified from self-directed delivery of constraint-induced movement therapy, electrical stimulation and therapy programmes that increase practice without using additional technology.