Purpose This study aims to systematically review the ethical challenges in artificial intelligence (AI)-enabled human resource management (HRM) and advance a legitimacy-based framework that explains how fairness, accountability and governance jointly shape stakeholder evaluations of algorithmic people decisions. Design/methodology/approach This review follows the scientific procedures and rationales for systematic literature reviews protocol and analyzes 87 Scopus-indexed articles published in ABDC 2022 A* and A journals. The analysis combines theories–contexts–characteristics–methods mapping, thematic synthesis and quality/risk-of-bias appraisal to distinguish descriptive patterns from deeper conceptual gaps. This procedure enables the review to identify not only what the literature counts but also how fairness, accountability and governance have been theorized, operationalized and empirically examined. Findings The literature is concentrated in recruitment and selection and remains dominated by organizational justice, trust, technology acceptance and algorithm aversion lenses. The synthesis identifies three recurring ethical tensions – objectivity versus embedded bias, efficiency versus procedural dignity and automation versus accountable human oversight – and shows that legitimacy depends on whether AI-supported HRM decisions are explainable, contestable, auditable and institutionally defensible. Practical implications This review provides guidance for aligning AI-enabled HRM governance with decision risk. High-stakes applications, including hiring, appraisal, promotion, compensation, monitoring and termination, require bias testing, accountability allocation, explainability protocols, human-review thresholds, appeal mechanisms, audit trails, vendor controls and mechanisms for employee and applicant voice. Originality/value This paper advances prior AI-HRM reviews by formally specifying a multilevel legitimacy framework and by treating fairness, accountability and governance as interdependent legitimacy conditions rather than separate ethical topics. It responds directly to calls for stronger sociotechnical, institutional and posthuman governance theorizing in algorithmically governed organizations.
Qadri et al. (2026) studied this question.