AI is remaking hiring: academic literature shows promise for faster, more objective recruitment but stops short of strong empirical proof; sentiment across studies is cautiously positive and calls for validation, oversight and real-world impact evaluation.

Harnessing AI for smarter hiring: machine learning and data analytics in transforming HR recruitment

Saeed Rouhani, Roshanak Aghabaghery, Hannan Amoozad Mahdiraji, Demetris Vrontis · September 03, 2026 · EuroMed Journal of Business

openalex review_meta n/a evidence 7/10 relevance Summary only summary available; pdf_status=paywall DOI Source PDF

Structured author observations

Linked only from stored provider relations; the raw author line above is never matched by name.

OpenAlex

Latest observation: September 8, 2026

Saeed Rouhani provider ID
Roshanak Aghabaghery provider ID
Hannan Amoozad Mahdiraji provider ID
Demetris Vrontis provider ID

A systematic review of 155 Scopus-indexed studies (2014–2024) finds AI is reshaping recruitment across five domains with cautiously positive but near-neutral lexical sentiment, while empirical validation and causal evidence about effects on hiring outcomes remain limited.

Citation observations

Cumulative provider counts captured on specific dates; providers are never combined.

0 cumulative citations

OpenAlex · Observed September 8, 2026

View corpus context

0 cumulative citations

Semantic Scholar · Observed September 8, 2026

View corpus context

Purpose Human resource (HR) departments increasingly face the complex task of identifying candidates who best align with organisational needs while reducing recruitment costs and time-to-hire. Traditional manual sourcing of applicants across multiple platforms, such as job portals and social media, has become inefficient in the era of data abundance. This study explores how artificial intelligence (AI) technologies, particularly machine learning (ML) and data analytics, are transforming recruitment and selection processes by enabling more objective, efficient and data-driven decision-making. Design/methodology/approach A systematic literature review (SLR) was conducted to map the intellectual structure of AI-driven recruitment research. From an initial pool of 1,444 publications indexed in Scopus, 155 met the inclusion criteria after rigorous screening. Latent Dirichlet allocation (LDA) topic modelling was applied to identify dominant research themes. In contrast, VADER sentiment analysis was applied to representative topic terms to generate exploratory weighted lexical-valence indicators and examine how evaluative language associated with these topics varied over time. Findings The analysis identified twenty themes, consolidated into five overarching research domains: (1) Skill-based assessment, (2) AI algorithms and techniques, (3) Candidate sourcing, (4) Industry-specific employment and (5) Employee lifecycle challenges. The topic-term sentiment results suggest a generally positive but cautious lexical-valence pattern in AI-enabled recruitment research between 2014 and 2024. However, because most scores were close to neutral under standard VADER thresholds, these findings are interpreted as exploratory indicators of evaluative language associated with topic terms rather than direct measures of authors' attitudes, contextual sentiment toward individual studies, or practical effectiveness of AI recruitment systems. Originality/value This study extends the understanding of AI adoption in HRM by combining topic modelling and topic-level sentiment analysis to reveal emerging trends, research gaps and practitioner implications. It contributes to both academia and practice by illuminating how ML and data analytics can be strategically leveraged to enhance fairness, efficiency, and predictive accuracy in recruitment, while also identifying areas where stronger validation, contextual analysis and human oversight remain necessary.

Summary

Main Finding

AI (ML and data analytics) is reshaping recruitment by enabling more objective, efficient and data-driven hiring decisions. A systematic review of the literature (2014–2024) identified 20 research topics grouped into five domains—skill-based assessment, AI algorithms and techniques, candidate sourcing, industry-specific employment, and employee lifecycle challenges—and finds the scholarly discourse to be cautiously positive about AI-enabled recruitment, though evaluations remain close to neutral and require stronger validation and human oversight.

Key Points

Scope and sample: 1,444 Scopus-indexed publications screened, 155 studies met inclusion criteria.
Methods: Combined systematic literature review with Latent Dirichlet Allocation (LDA) topic modelling and VADER sentiment analysis applied to topic terms.
Themes: 20 topics condensed into five domains:
Skill-based assessment
AI algorithms and techniques
Candidate sourcing
Industry-specific employment
Employee lifecycle challenges (onboarding, retention, development)
Sentiment: Topic-term sentiment shows a generally positive but cautious lexical-valence across 2014–2024; most scores hover near neutral under standard VADER thresholds, so results are exploratory indicators of evaluative language rather than measures of authors’ attitudes or system effectiveness.
Contributions: Integrates topic modelling and topic-level sentiment to surface trends, research gaps, and practitioner implications; highlights potential of AI to improve fairness, efficiency and predictive accuracy in hiring while emphasizing risks and validation needs.
Limitations: Sentiment analysis of topic terms is not a substitute for contextual, study-level evaluation; further empirical validation of AI tools in real-world hiring remains necessary.

Data & Methods

Data source: Scopus-indexed literature, initial N = 1,444 publications; final sample N = 155 after inclusion/exclusion screening.
Time window: Studies published between 2014 and 2024.
Analytical methods:
- Systematic Literature Review (SLR) to map the research field and apply inclusion criteria.
- Latent Dirichlet Allocation (LDA) topic modelling to identify dominant research themes (20 topics).
- VADER sentiment analysis applied to representative topic terms to compute weighted lexical-valence indicators and examine temporal variation in evaluative language.
Interpretation caveats: VADER results are treated as exploratory, focusing on language associated with topics rather than sentiment of full texts or empirical performance claims.

Implications for AI Economics

Productivity and cost effects: AI-enabled sourcing, screening and skill assessment can reduce time-to-hire and recruitment costs, raising recruiter productivity and potentially lowering hiring frictions across labor markets.
Matching efficiency and wages: Improved predictive accuracy and skill-based assessment may enhance worker–firm matching, affecting vacancy durations, job-to-job transitions, and potentially wages for better-matched candidates.
Distributional and fairness concerns: Algorithmic screening can entrench biases if models are poorly validated; addressing fairness requires investment in validation, explainability, and human oversight—costs that affect adoption decisions and the social returns to AI in HR.
Investment and adoption risk: Near-neutral/ cautious scholarly sentiment and identified gaps (validation, contextual analysis) indicate technological and regulatory uncertainty, which can slow diffusion and change expected returns on AI investments in HR systems.
Sectoral heterogeneity: Industry-specific findings imply heterogeneous adoption and labor-market impacts—some sectors may gain bigger productivity or matching benefits than others, influencing sectoral labor demand and capital–labor complementarities.
Policy and governance: Findings support policies promoting transparency, independent validation, and standards for algorithmic fairness in hiring to unlock efficiency gains while mitigating negative distributional effects.
Research and evaluation needs: For accurate economic assessment, future work should quantify effects of AI hiring tools on hiring costs, time-to-hire, match quality, turnover, and long-run earnings using field experiments and administrative worker–firm datasets.

Assessment

Paper Typereview_meta Evidence Strengthn/a — This is a systematic literature review and topic/sentiment analysis, not a primary empirical study that estimates causal effects; the paper synthesizes heterogeneous secondary evidence and provides exploratory lexical sentiment indicators rather than direct causal estimates. Methods Rigormedium — The study applies a transparent SLR screening (1,444 → 155), LDA topic modelling, and VADER sentiment on topic terms — standard and appropriate methods for mapping literature and lexical tone — but analytic choices (number of topics, LDA parameterization), reliance on topic-term rather than full-text/context sentiment, and heterogeneity in the underlying studies limit inferential strength and interpretation. SampleScopus-indexed academic literature published 2014–2024. Initial corpus N = 1,444 records screened; final sample of N = 155 studies meeting inclusion criteria; analyses use extracted topic-term representations from these studies for LDA and VADER lexical sentiment. Themeslabor_markets adoption productivity GeneralizabilityLimited to Scopus-indexed (academic) publications — excludes gray literature, industry reports, and non-indexed conference materials which may contain practitioner evidence., Potential language/publication bias (likely English-dominant and peer-reviewed journals/conferences)., Topic modelling (LDA) and lexicon sentiment (VADER on topic terms) are sensitive to preprocessing and modelling choices and do not capture full contextual meaning, limiting claim about authors' attitudes or real-world tool performance., Heterogeneity in quality, setting, methods, and outcomes across included studies reduces ability to generalize quantitative implications for matching, wages, or productivity., Findings map discourse and research gaps rather than estimate causal impacts or sector-specific effect sizes.

Claims (9)

Claim	Direction	Outcome	Confidence & Evidence	Details
The review screened 1,444 Scopus-indexed publications and included 155 studies after applying inclusion and exclusion criteria. Other	mixed	Scope and composition of the research literature	Reading fidelity high Study strength medium	n=1444 0.24
The literature review identified 20 research topics concerning AI-enabled recruitment and condensed them into five domains: skill-based assessment, AI algorithms and techniques, candidate sourcing, industry-specific employment, and employee lifecycle challenges. Other	mixed	Research-topic structure in AI-enabled recruitment	Reading fidelity high Study strength medium	n=155 0.24
The scholarly discourse on AI-enabled recruitment is generally positive but cautious, with topic-term sentiment scores tending to remain close to neutral. Other	positive	Lexical evaluative sentiment associated with AI-recruitment research topics	Reading fidelity high Study strength low	n=155 0.12
The sentiment results should be interpreted as exploratory indicators of evaluative language rather than as measures of authors' attitudes or the actual effectiveness of AI recruitment systems. Ai Safety And Ethics	mixed	Validity and interpretability of sentiment-based evaluation	Reading fidelity high Study strength high	n=155 0.4
The reviewed literature highlights the potential for AI-enabled recruitment to improve fairness, efficiency, and predictive accuracy in hiring. Decision Quality	positive	Hiring fairness, recruitment efficiency, and predictive accuracy	Reading fidelity high Study strength low	n=155 0.12
The literature identifies risks that algorithmic screening can entrench bias when recruitment models are poorly validated. Ai Safety And Ethics	negative	Risk of discriminatory or biased hiring decisions	Reading fidelity high Study strength low	n=155 0.12
The review concludes that stronger empirical validation and human oversight are needed before drawing firm conclusions about AI-enabled recruitment. Governance And Regulation	mixed	Validation and governance of AI-assisted hiring decisions	Reading fidelity high Study strength high	n=155 0.4
The review indicates that research on AI-enabled recruitment spans employee lifecycle challenges beyond initial hiring, including onboarding, retention, and employee development. Turnover	mixed	Coverage of onboarding, retention, and employee development in the research literature	Reading fidelity high Study strength medium	n=155 0.24
The review identifies industry-specific employment as a distinct domain, implying that the research agenda and potential effects of AI-enabled recruitment vary across sectors. Task Allocation	mixed	Sectoral variation in AI-enabled recruitment research and effects	Reading fidelity high Study strength medium	n=155 0.24