Artificial Intelligence (AI) in elementary language education: A systematic review

Synthesis: This PRISMA-guided systematic review asks what AI actually does in elementary Language Learning, and the answer is narrower than the field's rhetoric suggests. Across 31 studies published between 2013 and 2025, the work clusters on speaking, literacy and vocabulary, while grammar, listening comprehension and sign language are barely touched. Conversational AI and natural-language processing are common; adaptive systems and biometric feedback are not. Most studies are qualitative or mixed-methods with medium samples and no grade-level specificity, so the review's main contribution is a map of where the evidence is thin — and a call for developmentally appropriate, ethically grounded design rather than more undifferentiated AI-in-education claims.

Key Findings

1. Thirty-one peer-reviewed studies met inclusion criteria across six databases for 2013-2025, screened with dual independent reviewers and an inter-rater reliability of κ = 0.75.
2. The literature concentrates on speaking, literacy and vocabulary; grammar, listening comprehension and sign language are markedly underrepresented.
3. Pedagogical uses of AI centered on motivation, assessment and personalized interaction rather than on foundational skill development.
4. Conversational AI and natural-language processing were the most common technologies, while adaptive learning systems and biometric feedback were underused.
5. Most included studies used qualitative or mixed methods with medium sample sizes and lacked grade-level specificity, limiting what can be concluded about early-childhood learning.
6. The review is organized by the Diffusion of Innovation Theory and follows PRISMA and PICOS, with quality appraisal and risk-of-bias assessment included.

How the review was built

The review follows PRISMA and the PICOS framework, searching six major databases for peer-reviewed studies published between 2013 and 2025 and guided by the Diffusion of Innovation Theory. Thirty-one studies met the inclusion criteria and were analyzed through quantitative content analysis and thematic synthesis. Bias controls include dual independent screening, an inter-rater reliability of κ = 0.75, and a quality appraisal using the Mixed Methods Appraisal Tool.

Two research questions organize the analysis: which language-education aspects AI research addresses in elementary settings, and which AI technologies and applications predominate and how they support learning outcomes. The framing matters because the review is explicit that the elementary level is under-theorized — much of the evidence base generalizes from higher education or from undifferentiated "education" settings rather than studying young learners directly.

What AI supports, and what it leaves alone

The concentration of the literature is the finding. Speaking, literacy and vocabulary dominate, which aligns with where speech recognition and text generation are easiest to deploy. Grammar, listening comprehension and sign language are thin, and the authors treat that as a gap in both research and product design rather than a settled judgment about where AI belongs.

On the technology side, conversational agents and natural-language processing are the workhorses, supporting motivation, assessment and personalized interaction. Adaptive learning systems and biometric feedback appear far less often, despite being frequently promised in the broader AI-in-education literature. The mismatch between what is promised and what elementary classrooms actually study is one of the review's practical contributions.

The methodological gap

The review's second contribution is a critique of the evidence itself. Most included studies are qualitative or mixed-methods with medium samples, and many lack grade-level specificity — so a study labeled "elementary" may span ages where the developmental demands differ sharply. Without that specificity, claims about foundational language acquisition are hard to support.

The authors argue for developmentally appropriate, pedagogically coherent and ethically sound approaches, and they direct the methodological ask at researchers: stronger designs, clearer grade bands and attention to the overlooked skill areas. For policymakers the message is about infrastructure, Ethics and equity, since access to AI-supported language learning is unevenly distributed.

What this means for practice

• Instructors. Check what an AI language tool actually trains before adopting it. This literature concentrates on speaking, literacy and vocabulary, so a tool marketed for "language learning" may do nothing for grammar or listening comprehension.
• Curriculum designers. Specify the grade band when evaluating AI tools for early learners. The review finds most studies skip grade-level specificity, so evidence gathered on older children should not be assumed to transfer to early-childhood classrooms.
• Policymakers. Treat equity and infrastructure as part of adoption. The review frames ethical, inclusive and equitable use as a precondition rather than an afterthought, which matters where AI-supported provision is unevenly available.
• Researchers. Fill the named gaps directly — grammar, listening comprehension, sign language — with stronger designs and explicit grade bands, rather than adding another undifferentiated study of speaking practice.

Limitations

• The review includes 31 studies from a 2013-2025 window and six databases, so its map is bounded by the search strategy and by what was indexed.
• Most included studies are qualitative or mixed-methods with medium samples and no grade-level specificity, so the review can describe the field's shape but not pool effect sizes.
• The review is organized by a single theoretical lens, the Diffusion of Innovation Theory, which foregrounds adoption over learning mechanisms.
• Language coverage is not fully reported, so conclusions about multilingual and low-resource settings rest on a literature the review does not claim is representative.

Citation

Hamasha, M., Boulhrir, T., & Tiernan, P. (2026). Artificial Intelligence (AI) in elementary language education: A systematic review. Computers and Education Open, 11, 100424.