BackgroundArtificial intelligence (AI) decision aids can offer individualised and tailored information to patients and clinicians to inform shared decision making (SDM). We aimed to explore the perspectives of clinicians and patients on the use of AI decision aids to inform SDM in different clinical settings. MethodsWe did a systematic review by searching four large databases (MEDLINE, CINAHL, SCOPUS, and Embase), using appropriate MeSH terms and key words (“perceptions”, “clinicians”, “patients”, “artificial intelligence”, “decision-aids”, “shared decision-making”), for articles published in English from database inception until Oct 5, 2020. All quantitative, qualitative, mixed-method studies that discussed the perspectives of clinicians or patients from any medical speciality, clinical setting, and timeperiod were included and reviewed by two independent reviewers (inter-rater reliability was 100%). Studies discussing decision aids without an AI component were excluded. We assessed quality and bias using the Critical Appraisal Skills Programme (CASP) checklist. We did a narrative synthesis based on the Research Methods Programme framework developed by the UK Economic and Social Research Council. This review was registered with PROSPERO (number CRD42020220320) and adhered to the PRISMA guidelines. Findings509 articles were identified, with 17 articles eligible for inclusion. All included studies received a score of 7 or higher on the CASP checklist. AI decision aids were used to support SDM at various stages of the patient journey, including prevention, screening, prognosis, diagnosis, and treatment. It was perceived by patients and clinicians that these decision aids helped to promote patient engagement and communication with their clinicians, increase patient confidence and compliance, inform the consenting process, present risk estimates in an interactive and individualised manner, and improve patient satisfaction around their overall clinical care. The main barriers to using such aids were around patients' variability in technology literacy, and incomplete or missing information that could potentially affect the findings. InterpretationAI decision aids were perceived to be useful tools to inform the SDM process for both patients and clinicians at different stages of the patient journey. However, some AI decision aids were used in a simulated research environment, and so clinicians' and patients' perspectives on these might differ if used in the clinical setting. FundingNewcastle University Overseas Research Scholarship.