Bridging the Gap: Investigating Students' Self-Reported AI-Literacy and Perceived Employability Competencies for an AI-Enabled Workforce

Synthesis: This cross-sectional survey examined how self-reported AI literacy relates to perceived employability competencies in 380 university students recruited through Prolific, pooled into Middle East (n=284; 74.7%) and Gulf (n=96; 25.3%) residence codes. The contextually compiled SAILEC-2025 questionnaire retained four self-reported AI literacy dimensions (Understand & Knowledge, Analyze & Evaluate, Attitudes & Mindsets, and a combined Use & Application/Create & Innovate composite) and eight perceived employability composites after exploratory factor analysis. Mean self-ratings were high in both domains. Hierarchical regression showed that the four AI literacy dimensions added ΔR2=0.554 to a model explaining 60.8% of the variance in overall perceived employability readiness, with Analyze & Evaluate strongest. In PLS-SEM, Attitudes & Mindsets was statistically involved in theory-ordered indirect associations linking Understand & Knowledge and Use & Application/Create & Innovate to all eight employability outcomes. Every measure was a single-session student self-assessment, so no causal, objectively assessed competence, employer-rated, or employment-outcome claims follow.

Key Findings

• Self-reported AI literacy means were uniformly high, from M=4.17 (SD=0.64) for Analyze & Evaluate to M=4.57 (SD=0.58) for Attitudes & Mindsets.
• Perceived employability composites ranged from M=4.14 (SD=0.69) for Adaptability to M=4.57 (SD=0.48) for Learning Agility & Professional/Ethical Behavior.
• Five background controls explained R2=0.054 of the variance in overall perceived employability readiness; adding the four AI literacy dimensions produced ΔR2=0.554 (p<0.001).
• In the final readiness model all four dimensions were significant: Analyze & Evaluate β=0.286, Attitudes & Mindsets β=0.266, Use & Application/Create & Innovate β=0.217, Understand & Knowledge β=0.187 (all p<0.001).
• For career readiness only Analyze & Evaluate (β=0.216, p<0.001) and Attitudes & Mindsets (β=0.192, p=0.004) were significantly associated (ΔR2=0.263; R2=0.280).
• It was predicted by Use & Application/Create & Innovate (β=0.511) and Understand & Knowledge (β=0.276) but not by Analyze & Evaluate (β=0.046, p=0.344); indirect associations through it were significant for those two across all eight outcomes.

Design, instrument, and sample

SAILEC-2025 was contextually compiled for students in Middle East and Gulf Higher Education settings rather than previously validated, with items generated deductively from established AI literacy and employability frameworks and reviewed by two subject-matter experts. Of 33 AI literacy and 49 employability items administered, 30 and 46 were retained. An exploratory factor analysis (principal axis factoring, oblimin rotation) in SPSS 31 preceded reflective measurement assessment in SmartPLS 4.0 on the same 380 respondents. Internal consistency was high: total AI literacy α=0.944, with every subscale between α=0.772 and α=0.912, and employability composites α=0.966. Use/application and create/innovate items overlapped empirically and were merged, so no separate associations are attributed to them. Data were collected in October 2025. The sample skewed male (71%), STEM-enrolled (67.4%), and AI-experienced (90.3% frequent users).

Self-rated levels and their associations

Self-reported AI Literacy was positively associated with every perceived employability construct. The strongest bivariate association was between Attitudes & Mindsets and Use & Application/Create & Innovate (r=0.716), followed by Understand & Knowledge and Use & Application/Create & Innovate with digital and technology literacy (r=0.631 and r=0.624), and Analyze & Evaluate with Critical Thinking (r=0.602). Because every construct came from the same respondents in one session on the same five-point scale, with no temporal or source separation, shared method variance could not be eliminated; the uniformly high means also leave ceiling effects and socially favorable self-reporting as plausible explanations. The paper treats these levels as descriptions of how students in these settings understand their preparedness, not as demonstrated capability.

Regression and indirect-association results

Two hierarchical regressions tested the dimensions against overall perceived employability readiness (a 46-item mean) and perceived Career Development and Readiness. The full readiness model accounted for 60.8% of the variance, with Analyze & Evaluate strongest (β=0.286) and no control variable significant. The career-readiness model reached R2=0.280, with only Analyze & Evaluate (β=0.216) and Attitudes & Mindsets (β=0.192) associated; the confidence intervals for the other two dimensions included zero. In the PLS-SEM model, Attitudes & Mindsets was specified as an intervening affective-orientational dimension; Fit was modest (NFI 0.684 saturated, 0.622 estimated). Indirect associations through it were significant for Use & Application/Create & Innovate (β=0.229 to β=0.337) and Understand & Knowledge (β=0.124 to β=0.182) across all eight outcomes, and none for Analyze & Evaluate. Because direct antecedent paths were omitted, mediation was not tested. The Technology Adoption Models and social cognitive career frameworks served only as interpretive lenses; perceived usefulness, ease of use, behavioral intention, and task-specific Self-Efficacy were not measured.

What this means for practice

• Treat AI literacy as multidimensional: the evaluative (Analyze & Evaluate) and affective (Attitudes & Mindsets) dimensions behaved differently from foundational knowledge in this sample.
• Where programs target perceived career readiness, attend to critical evaluation of AI outputs and students' orientation towards AI; knowledge and hands-on use were weaker there.
• Do not read high self-ratings as competence: pair self-report instruments with objective knowledge tests and authentic AI-task performance assessments, and report sample composition alongside results, before drawing program conclusions.

Limitations

• All focal variables came from one anonymous, single-wave, same-source questionnaire, so it measures perceptions rather than demonstrated competence, employer judgments, or employment outcomes, and cannot establish temporal or causal order.
• SAILEC-2025 had not previously been validated in these contexts, and the factor analysis and SmartPLS assessment used the same sample, so the retained structure is not independently confirmed; two employability pairs exceeded the strict HTMT criterion of 0.85.
• Country of residence was retained only as two pooled codes, measurement invariance was not tested, and no country-level or institutional data were available, so results cannot be attributed to particular national systems.
• The sample was predominantly male (71%), STEM-enrolled (67.4%), frequent AI users (90.3%), and work-experienced (86.1%); the post hoc sensitivity analysis could not detect effects as small as Cohen's f2=0.02.

Citation

ElSayary, Areej; Ragab, Karim. (2026). Bridging the Gap: Investigating Students' Self-Reported AI-Literacy and Perceived Employability Competencies for an AI-Enabled Workforce. Journal of Computer Assisted Learning, 42, e70327. https://doi.org/10.1002/jcal.70327