Background: Atrial fibrillation (AF) is an important risk factor for ischemic stroke but has no recognized screening method. Wearable devices have the potential to provide near continuous monitoring to detect AF. This systematic review evaluates the current state of software capable of detecting AF using wearable devices. Methods: We conducted a systematic search using PRISMA method of Medline, CENTRAL, PubMed and trial registries up to January 15, 2017. Abstracts and titles were screened, and relevant articles reviewed fully. English articles were selected if reporting on (1) software for AF detection (2) using heart rhythm signal, (3) theoretically applicable to wearable technology. Quality was evaluated with Cochrane GRADE. Results: Of 269 unique abstracts, 54 were identified for full review. 20 studies met inclusion criteria for algorithm accuracy analysis. Sensitivity and specificity ranged from 87.0 - 97.6% and 89.0 - 99.6%, respectively. 4 studies analyzed signal acquired using mobile devices with similar accuracy. Algorithms were potentially portable to wearable devices. Qualitative observations on the state and applicability of technology were made. Conclusions: Software analysing heart rhythm may be accurate for AF screening, but has not been tested on wearable devices. Such technology is promising but may be limited by hardware accuracy and high false positive rates.