Artificial Intelligence in Recruitment: Efficiency, Algorithmic Bias, and the Future of Human Resource Selection

Abstract

Artificial intelligence is increasingly used across recruitment and personnel selection to source candidates, parse résumés, match applicants with vacancies, communicate through chatbots, summarize interviews, and recommend hiring decisions. These systems promise improvements in speed, scalability, consistency, and administrative efficiency. Employment, however, is a particularly consequential application because selection decisions determine access to income, career development, and economic opportunity. This paper examines whether AI-based recruitment can improve hiring efficiency without reproducing or amplifying human discrimination. Evidence indicates that AI can automate high-volume administrative activities and standardize candidate evaluation, but standardization should not be confused with neutrality. Audits of language models and embedding systems show that demographic signals can alter résumé ranking even when job-related information remains constant, and early comparative evidence suggests that the direction of such bias can differ substantially between model generations. Candidate reactions also complicate the business case because algorithmic decision-making is often associated with weaker perceptions of justice, trust, organizational attractiveness, and job-pursuit intentions. The paper argues that the most defensible architecture is structured human-AI collaboration. AI should automate repetitive and information-intensive recruitment tasks, while validated assessments, independent bias testing, explainability, human review, appeal mechanisms, and legal accountability remain central to consequential employment decisions. The European Union AI Act reflects this risk-based approach by treating many recruitment and candidate-evaluation systems as high-risk.