Transformations in Talent Acquisition: Measuring AI’s Effects on Recruitment Efficiency and Bias

2.1. AI and recruitment efficiency

The promise of AI in recruitment is often centered around gains in operational efficiency. Tools such as machine learning–driven résumé parsers, predictive analytics for job matching, and automated scheduling systems are credited with reducing time-to-hire and administrative burden (Upadhyay & Khandelwal, 2018; Meijerink et al., 2021). According to Chamorro-Premuzic et al. (2019), companies implementing AI-enabled screening systems reported up to 40% faster shortlisting processes, especially in high-volume hiring contexts.

However, these improvements are not uniform. As noted by Liem et al. (2018), the effectiveness of AI varies depending on dataset quality, job specificity, and recruiter adaptability. Moreover, efficiency does not always translate into accuracy systems trained on outdated data or biased hiring histories may replicate inefficiencies rather than resolve them.

2.2. Candidate experience and perception

Candidate experience is another area reshaped by AI technologies. Chatbots, automated emails, and self-service platforms aim to create smoother, faster interactions. Research shows that applicants appreciate timely responses and clear communication, which AI tools can help deliver at scale (Suen et al., 2019; Dastin, 2018). Nevertheless, perceived fairness and transparency play a central role in whether candidates trust automated systems.

Recent studies have shown that applicants often feel alienated when recruitment decisions are made by “black-box” systems. Langer et al. (2021) argue that while digitalization offers operational benefits, it may erode the human connection valued in recruitment, particularly in early career or sensitive roles.

2.3. Algorithmic bias and fairness in hiring

Perhaps the most contested dimension of AI in recruitment is algorithmic bias. Studies have documented how AI tools can perpetuate or even exacerbate existing inequalities, especially in terms of gender, ethnicity, and disability (Binns et al., 2018; Raghavan et al., 2020). A notable example is Amazon’s abandoned recruitment engine, which was shown to penalize CVs that included words associated with women (Dastin, 2018).

These risks have spurred calls for stronger auditing and fairness-by-design approaches. Zhang and Dafoe (2020) advocate for “algorithmic hygiene”—regular monitoring and bias mitigation protocols—to avoid unintended discrimination. Despite these initiatives, the lack of industry-wide standards and the proprietary nature of many algorithms hinder transparency and oversight.

2.4. Transparency, trust, and ethical governance

The trustworthiness of AI-based recruitment systems hinges on their transparency and explainability. Yet many commercially available tools offer limited insight into how decisions are made. This lack of explainability is particularly problematic in jurisdictions with emerging data protection and anti-discrimination laws (Binns et al., 2018; Mehrabi et al., 2021).

Research by Chandad and Abakouy (2025) illustrates how perceptions of fairness and psychological safety are crucial for employee well-being in industrial contexts. Their findings highlight that opaque systems whether human or algorithmic tend to erode trust, particularly among underrepresented groups. Consequently, the deployment of AI in recruitment cannot be seen as a purely technical exercise; it must also be understood as a cultural and ethical challenge within organizations.

3.1. Research design

This study adopts a quantitative cross-sectional research design, aimed at examining the impact of AI-based recruitment systems on key performance indicators in talent acquisition. We developed a structured survey instrument to capture HR professionals’ perceptions and experiences regarding the use of AI in their recruitment processes. The study follows ethical standards for data collection and analysis and complies with reproducibility standards.

3.2. Sample and data collection

The sample consists of 423 human resource professionals from diverse industries, including manufacturing, technology, financial services, and logistics. Participants were selected using a stratified purposive sampling method to ensure representation across firm sizes (SMEs and large enterprises), geographical regions (Europe, North Africa, and the Gulf ), and AI adoption levels (pilot vs. fully deployed).

Data were collected between January and March 2025 via a secure online survey platform. Participation was voluntary and anonymous, with informed consent obtained electronically prior to the start of the survey.

3.3. Survey instrument

The questionnaire comprised 28 closed-ended items organized into four dimensions:

Responses were captured using a five-point Likert scale (1 = strongly disagree to 5 = strongly agree). The instrument was pre-tested on 15 HR professionals to ensure clarity, internal coherence, and face validity.

3.4. Data analysis

We conducted descriptive statistics, exploratory factor analysis (EFA), and multiple linear regression models to identify patterns and assess the relationship between AI usage and perceived recruitment outcomes. Cronbach’s alpha was used to test internal reliability for each construct (α > 0.80 for all dimensions).

Regression diagnostics were performed to test for multicollinearity, normality, and heteroscedasticity. Statistical analyses were carried out using SPSS v28 and R (v4.3).

4.1. Sample characteristics

The final dataset included 423 valid responses. The majority of participants (62%) held managerial or senior HR roles, with 38% representing operational or mid-level positions. Regarding company size, 54% worked in firms with over 500 employees, whilst 46% were from small and medium enterprises (SMEs). AI-based recruitment systems were in full deployment in 39% of the companies surveyed, and at pilot or testing stages in 61%.

Table 1 summarises the demographic and organisational characteristics of respondents.

4.2. Descriptive statistics

Respondents generally reported positive perceptions of AI’s impact on recruitment.

4.3. Exploratory Factor Analysis (EFA)

An exploratory factor analysis (principal axis factoring, Varimax rotation) confirmed the validity of the four predefined dimensions.

4.4. Regression analysis

We ran multiple linear regressions to assess the impact of AI tool usage (independent variable) on each recruitment dimension (dependent variables). Control variables included company size, industry sector, and geographic location.

Key findings:

See Table 2 for full regression results and Figures 1–3 for visual summaries of the effects by job category and AI maturity stage.

Figure 1. Perceived impact of AI on recruitment dimensions (mean scores).

Mean scores across recruitment dimensions with standard deviation error bars (5-point Likert scale).

Figure 2. Regression coefficients of AI impact on recruitment outcome.

Standardised regression coefficients showing the impact of AI usage on recruitment outcomes. Significance levels: *** p < 0.001; ** p < 0.01; * p < 0.05; n.s.: Not significant.

Figure 3. Distribution of survey responses across participants.

Distribution of responses across recruitment dimensions (N = 423).

6.1. Conclusion

This study provides one of the few quantitative analyses to date on the effects of Artificial Intelligence in recruitment processes. Based on a multi-sector survey of 423 HR professionals, the findings demonstrate that AI tools significantly improve recruitment efficiency and contribute to a more structured candidate experience. However, these technological gains are accompanied by persistent concerns related to fairness, transparency, and the risk of reproducing systemic biases.

Our results highlight that AI, whilst effective in automating certain phases of hiring, does not inherently guarantee equity or trust. These dimensions depend not only on the algorithmic design but also on the broader organisational culture, regulatory frameworks, and ethical safeguards put in place. As emphasised in the literature (Binns et al., 2018; Chandad & Abakouy, 2025), meaningful adoption of AI in HR must go beyond operational logic and engage with deeper issues of accountability and inclusion.

From a managerial perspective, these findings suggest that AI should be integrated into recruitment as part of a hybrid human–machine decision-making model, where human oversight remains central. Developers, HR practitioners, and policymakers should collaborate to co-design tools that are both technically robust and ethically defensible.

6.2. Limitations

While this study offers valuable insights, several limitations must be acknowledged:

6.3. Future research

Future studies could address these limitations by adopting longitudinal designs, enabling an understanding of how perceptions and outcomes evolve over time with AI implementation. Moreover, experimental studies comparing traditional and AI-assisted recruitment pipelines could offer stronger causal evidence on performance differences.

Additional research is also needed on the candidate perspective, particularly among underrepresented groups, to assess how AI shapes their trust, comfort, and perceived fairness throughout the hiring process. Finally, interdisciplinary work combining data science, labor law, and organizational behavior would enrich the field by bridging technical insights with human-centered values.