How AI literacy correlates with affective, behavioral, cognitive and contextual variables: A systematic review

Synthesis: Bewersdorff, Nerdel and Zhai synthesize the empirical evidence linking AI Literacy to surrounding variables, analyzing 31 studies that used six of the 15 instruments catalogued by Lintner (2024), covering 14 countries and a combined sample of N = 12,071. Structuring results with the affect, behavior, cognition framework extended by a background and surrounding category, they find consistently strong positive correlations with AI self-efficacy, positive AI attitudes, Motivation and digital competencies, and negative correlations with AI anxiety. Demographic variables barely move the needle: age, gender and socio-economic status show weak or non-significant correlations. The decisive caveat is measurement. Self-assessment scales produced substantially stronger correlations than the single performance-based test, with AI self-efficacy correlating at r = .84 versus r = .26 after weighting, which the authors read as Self-Efficacy and Assessment Validity entanglement, and possible metacognitive inflation, rather than as evidence about AI Literacy itself.

Key Findings

Thirty-one studies from 29 publications and six AI literacy instruments spanned 14 countries with a combined sample of N = 12,071 (median 300, range 113 to 1,465).
Motivation in AI-supported learning correlated positively, from 0.350 up to 0.71, with the highest value reported among teachers.
AI anxiety correlated negatively with AI literacy, from −0.19 to −0.645, while positive AI attitudes (GAAIS) correlated 0.47 to 0.82 and negative attitudes up to −0.58.
AI self-efficacy and self-competency correlations ranged from 0.14 to 0.93, and digital literacy constructs from 0.21 to 0.85.
After Fisher's z weighting, self-assessment scales yielded AI self-efficacy r = .84 versus r = .26 for the performance-based test, and positive AI attitude r = .53 versus r = .20.
Age correlated weakly or non-significantly with AI literacy (mean ages 21.4 to 33.3 years), and socio-economic status correlated 0.081 (family) and −0.089 (student).
Intention to use AI correlated 0.21 to 0.54 among university students and 0.76 among teachers, but the performance-based test showed 0.00 with AI use.

How the Review Was Built

The review follows PRISMA 2020, with identification and screening done by the authors without AI screening tools. It starts from Lintner's (2024) evaluation of 16 AI literacy instruments and retains 15 candidates. A Google Scholar "Cited by" search on February 3 and 4, 2025 returned 982 publications; duplicates left 698, non-English exclusion left 576, review exclusion 544, abstract access 489, and empirical-data restriction 359. Requiring correlations with a full AI literacy score, not subscales, narrowed this to 29 publications reporting 31 studies. Variables were grouped by systematic content analysis into affective, behavioral, cognitive and background or surrounding categories. The self-assessment versus performance-based comparison covered three variables and used Fisher's z transformation weighted by study sample sizes.

What Correlates With AI Literacy

Affective results are the most consistent. Higher AI literacy went with lower AI anxiety (−0.19 to −0.645) and more favorable attitudes: positive attitudes ranged 0.47 to 0.82 while negative attitudes reached −0.58. AI acceptance correlated 0.29 to 0.6, though a study of IT leaders reported a non-significant 0.06. On behavior, intention to use AI ranged 0.21 for Chatbot usage to 0.54 for continuance intention among students, and reached 0.76 for teachers. Reported use correlated positively in most studies (use at work 0.35, at school 0.29, prior use 0.49). On cognition, motivation ran 0.350 to 0.71, academic self-efficacy 0.473, innovative mindset 0.554, and digital or AI literacy 0.21 to 0.85. Interest in AI was the weak spot: 0.06 and 0.36, against 0.59 for actively seeking information about AI.

Where the Evidence Conflicts

Two findings resist synthesis. Trust in AI correlated positively with AI literacy in some studies (0.29 to 0.539) but not in others: perceived bias, plagiarism, fabrication and harm measures clustered between −0.06 and −0.09 without significance, and one IT leader sample reported −0.13. Second, the relationship between AI literacy and AI use splits by instrument. Self-assessment studies found positive correlations, whereas the performance-based test reported 0.00, which the authors read as a challenge to digital-native assumptions and as a definitional problem in how use is measured. Measurement choice also drives correlation strength: self-assessment instruments correlated 0.86 with another self-assessment scale but only 0.21 with the performance-based test. The authors conclude that self-assessment scales may partly capture self-efficacy rather than objective AI literacy.

What this means for practice

Instructors. Pair AI literacy instruction with affective and behavioral components, because anxiety, attitudes, motivation and intention all moved with literacy while age and background did not.
Instructors. Do not treat frequent AI use as evidence of literacy, since the performance-based test found 0.00 with AI use.
Researchers. Triangulate self-assessment with an objective measure before high-stakes decisions: the same construct correlates r = .84 with AI self-efficacy when self-assessed and r = .26 when tested.

Limitations

Only six of the 15 instruments identified by Lintner (2024) had been used in studies reporting correlations, and all were published after 2021.
Twenty of the 31 studies sampled university or college students, eight in medical and care disciplines; primary and secondary students are absent entirely.
Only one instrument in the review is performance-based and appeared in just two studies, so that comparison rests on a thin base.
The correlations are bivariate and not causal, and narrow age ranges (means 21.4 to 33.3 years) may hide age effects.

Connected Concepts

Connected Articles

How to Assess AI Literacy: Misalignment Between Self-Reported and Objective-Based Measures — How to Assess AI Literacy: Misalignment Between Self-Reported and Objective-Based Measures
The GenAI Skill Bypass: Mapping Divergent Pathways of University Students and Staff AI Literacy — The GenAI Skill Bypass: Mapping Divergent Pathways of University Students and Staff AI Literacy
Measuring Acceptance of Age-Tiered AI Literacy Guidebooks: A Developmentally Informed Study of K-12 Students and Teachers — Measuring Acceptance of Age-Tiered AI Literacy Guidebooks: A Developmentally Informed Study of K-12 Students and Teachers
Technology, Education and Critical Media Literacy: Potential, Challenges, and Opportunities — Technology, Education and Critical Media Literacy: Potential, Challenges, and Opportunities
AI Literacy Interventions in Education: A Meta-Analysis of Effects and Moderators — AI Literacy Interventions in Education: A Meta-Analysis of Effects and Moderators
Teacher education for artificial intelligence literacy through a self-determination theory perspective — Teacher education for artificial intelligence literacy through a self-determination theory perspective

Citation

Bewersdorff, A., Nerdel, C., & Zhai, X. (2025). How AI literacy correlates with affective, behavioral, cognitive and contextual variables: A systematic review. Computers and Education: Artificial Intelligence, 9, 100493.

Metadata

Study design	Literature Review
Level	K 12, Higher Ed
Intended audience	Instructors, Learners, Researchers
Learning and instruction	Motivation, Self-Efficacy, Anxiety and Stress
Assessment and measurement	Educational Measurement, Assessment Validity
Research methods and evaluation	Meta-Analysis and Systematic Review, Research Methods in AIED
Foundations of AI in education	AI Literacy, Educational Development

Source History Edit Chat