Governing Algorithmic Gender Bias in AI-Based Recruitment: A Systematic Evidence Map of Accountability, Oversight, and Responsible Deployment

AI-mediated recruitment is a consequential sociotechnical system in which local model choices can redistribute employment opportunity across organizations and labour markets. This systematic evidence map examines whether research on gender bias has progressed from disparity diagnosis to governance through mitigation, transparency, auditability, human authority, and risk-based compliance. Scopus and Web of Science searches yielded 173 records; 139 remained after de-duplication and 101 after broad screening. Fifty-eight candidates advanced to retrieval. Twenty-three full texts were recovered, one was excluded, and the final hybrid map comprised 57 records: 22 with full-text-supported eligibility and 35 with title/abstract-supported eligibility. For comparability, six governance dimensions were uniformly coded from indexed text. A disclosed AI-assisted consistency audit reapplied the rules to 101 eligibility decisions and 342 coding cells; it reproduced the structured dataset but was not an independent human review. Bias measurement appeared in 24/57 records (42.1%, 95% CI 30.2–55.0), technical mitigation and accountability/audit each in 9/57 (15.8%), and human oversight in 5/57 (8.8%). Only one record showed integrated governance coverage. No temporal comparison remained significant after multiple-testing correction. We propose a closed-loop systems model linking measurement, accountable monitoring, disclosure and contestability, and human oversight with adaptive re-validation. The synthesis estimates observable governance reporting, not intervention efficacy.