As AI literacy becomes a global educational priority, a growing number of interventions have been introduced across diverse contexts. However, it remains unclear how, and under what conditions, these interventions support student development in AI literacy. To address this gap, we conducted a three-level meta-analysis of 59 empirical studies, synthesizing 172 effect sizes drawn from 7,211 participants. The analysis estimates the overall impact of AI literacy interventions and examines variation across study characteristics, instructional approaches, and assessment methods. Results show a large and statistically significant overall effect size ( g = 0.837, p < .001). Notably, substantial heterogeneity and a wide 95% prediction interval -0.292, 1.966 suggest that the effectiveness of AI literacy interventions varies considerably across educational settings and may not generalize uniformly across contexts. Moderator analyses identify the geographical region and learning-outcome focus as significant factors: interventions in East Asia and Europe yielded stronger effects than those in North America, and knowledge-focused interventions outperformed those targeting skills, attitudes, or ethics. Despite non-significant moderator effects, larger effect sizes were observed for mixed or reflective pedagogies, generative AI tools, and performance-based tasks, warranting further investigation. Together, these findings suggest that AI literacy education should move beyond knowledge and conceptual understanding toward skills, practices, ethics, and attitudes. This shift may be supported by integrated pedagogies that combine conceptual learning, hands-on making, reflection, and GenAI-supported tools. The geographic variation also highlights the need for culturally relevant and context-sensitive interventions that respond to local educational priorities, technological conditions, and learner needs.
Liu et al. (Thu,) studied this question.