Rethinking Elementary Education's Writing Instruction in The Age of Generative AI: A Systematic Review

Synthesis: This systematic literature review synthesizes 8 peer-reviewed studies (2019–2025) on AI literacy for elementary writing instruction, finding that AI integration efficiently supports writing practices and fosters creativity through multimodal application while creating nuanced approaches to writing assessment — alongside unresolved limitations for future research. Across four thematic categories — writing practice, creative and multimodal writing, instructional scaffolding, and assessment — the review positions AI shifting from an efficiency tool to an integrated pedagogical resource supporting both cognitive and creative dimensions of writing development.

Key Findings

1. Eight studies, four themes. The review synthesizes 8 peer-reviewed studies (2019–2025) that fall into four thematic categories: AI for writing practice and performance, AI for creative and multimodal writing, AI as instructional scaffolding, and AI for writing assessment.
2. AI efficiently supports writing practice. Conversational AI tools (e.g., chatbot systems) produce significant gains in elementary students' writing performance and foster positive learner perceptions of the writing environment.
3. Creativity through multimodality. Generative AI that integrates visual, textual, and multimodal elements extends writing beyond text-based practice, improving writing disposition, motivation, and early literacy outcomes.
4. Nuanced assessment, with human oversight required. Automated and multimodal assessment systems show promise for efficient feedback, but human oversight remains essential for fairness, accuracy, and appropriate pedagogical use.

Overview and Research Question

Writing is a foundational component of literacy development in elementary education, supporting students' organization of ideas, communication of meaning, and interaction across academic disciplines while also promoting cognitive development through the exchange of ideas and creative meaning-making with language. Because early experiences shape long-term attitudes toward writing, they play a critical role in both academic achievement and lifelong learning. Yet writing remains the most challenging literacy skill for young learners to develop, requiring the coordination of multiple cognitive processes — constructing ideas, structuring information, selecting vocabulary, and applying grammar — simultaneously. This gap between the recognized importance of writing instruction and limited classroom time devoted to it motivates the search for innovative, scalable approaches.

The review addresses the question: How is artificial intelligence being used to support writing instruction in elementary education? It synthesizes systematic literature review findings from a dataset originally drawn from a broader review of AI in K-12 literacy instruction, re-screened with a narrower analytical focus on elementary writing contexts.

Method

Following established PRISMA guidelines for systematic reviews, the researchers searched IEEE Xplore, Scopus, Web of Science, ERIC, Wiley Online Library, ACM Digital Library, and Google Scholar using tailored Boolean strings targeting AI, writing instruction, and elementary/primary education. Records were limited to English-language peer-reviewed journal articles published between 2019 and 2025, exported to Zotero for deduplication, and screened against explicit inclusion and exclusion criteria focused on elementary-level students and/or teachers. Two researchers coded the retained studies inductively and independently in a shared spreadsheet, resolving discrepancies through iterative discussion and consensus, then conducted a thematic synthesis that produced four thematic categories. No formal inter-rater reliability statistic was calculated given the small corpus.

AI for Writing Practice and Performance Development

One included study (Kwon et al., 2023) examined a chatbot-based system for writing practice among elementary English language learners. Students in the experimental group engaged in writing activities with the chatbot for fifteen weeks, while the control group received traditional teacher-led instruction. Learners who used the chatbot demonstrated significantly higher post-test writing performance, and they reported positive perceptions of the chatbot learning environment. The interactive nature of conversational AI offered a comfortable and supportive space for practicing language and developing writing skills, indicating that such tools can serve as valuable supplements to classroom writing instruction by enabling repeated practice, interactive Feedback, and greater engagement with the writing process.

AI for Creative and Multimodal Writing

Generative AI models can produce images, text, and other digital media artifacts that serve as stimuli for storytelling and narrative development, expanding writing beyond traditional text-based practice through multimodal composition. Kaskaya and Ates (2025) found that AI-supported Visualization applications improved fourth-grade students' writing disposition — including confidence, persistence, and enthusiasm — and stimulated idea generation and imagination. Gultekin et al. (2025) similarly found that generative AI-assisted picture books enhanced motivation to participate in literacy activities and supported early literacy development such as phonemic awareness and reading fluency, while cautioning that teachers must critically evaluate the quality and instructional appropriateness of AI-generated materials, especially for students with reading or writing difficulties. The review also draws on eye-tracking work suggesting that young learners employ distinct cognitive strategies — prioritizing planning and rereading versus rapid action and visual feedback loops — which can inform how AI environments accommodate diverse student processing during multimodal composition.

AI as Instructional Scaffolding

Rather than functioning solely as a writing tool, AI can act as an instructional scaffold, supporting teachers in modeling genre structures, providing example texts, and guiding students' understanding of writing conventions. de Oliveira and dos Santos (2025) explored how generative AI produces mentor texts for genre-based second-language writing instruction, drawing on systemic functional linguistics to illustrate structural stages and linguistic features across genres. Ng et al. (2022) examined digital story writing as an inquiry-based pedagogy integrating AI Literacy with literacy instruction among primary students, finding that the inquiry-based learning cycle — orientation, conceptualization, investigation, conclusion, discussion — supported both writing development and understanding of AI. The review stresses that AI environments require tight alignment with intentional lesson sequences and teacher-guided scaffolds to move learners from knowledge building to abstract application, echoing instructional design research on intentional tool mediation and strategic prompting.

AI for Writing Assessment and Evaluation

Advances in machine learning have enabled automated systems that analyze multiple aspects of student writing, including organization, vocabulary, and task fulfillment. Hannah et al. (2023) examined the validity of machine learning models evaluating writing traits in grades three through six, finding reasonable agreement with human raters across traits like organization and vocabulary, but also limitations in detecting off-topic or nonsensical responses and in maintaining consistent scoring across student groups. Smith et al. (2019) proposed a multimodal assessment framework integrating writing and drawing as complementary indicators of student understanding. Wilson and Wen (2022) used automated essay scoring to show that upper-elementary students' metacognitive knowledge about writing predicted their writing scores across genres. Across these studies, the review concludes that although automated evaluation shows promise for efficient formative assessment, human oversight remains essential to ensure fairness, accuracy, and appropriate pedagogical use.

Cross-Cutting Concerns: Equity, Ethics, and Overreliance

Across all four themes, the reviewed studies reveal concerns for diverse and multilingual learners, including issues of differential access, overreliance on AI feedback, and ethical issues such as data privacy, transparency, and the quality of the AI tool itself. The review draws on scholarship emphasizing that mitigating overreliance requires AI tools to offer transparent and editable outputs, allowing learners to preserve cognitive ownership and exercise Learner Agency over their self-regulated learning processes.

What this means for practice

• Instructors. Use conversational AI as a supplement for repeated practice rather than a replacement for teacher-led writing time; the one writing-practice study in this corpus ran a chatbot condition for fifteen weeks against traditional instruction and reported significantly higher post-test writing performance and positive learner perceptions.
• Instructors. Add Multimodal AI composition tasks, such as AI-supported Visualization and generative picture books, to reach writing disposition and early literacy: the fourth-grade visualization study reports gains in confidence, persistence, and enthusiasm, and the picture-book study reports greater motivation, phonemic awareness, and reading fluency.
• Instructors. Evaluate AI-generated mentor texts and materials before classroom use for quality and instructional appropriateness, particularly for students with reading or writing difficulties, since the review warns that AI output can be pedagogically unsuitable even when fluent.
• Instructors. Treat automated writing scores as evidence to interrogate rather than verdicts: machine scoring agreed reasonably with human raters on traits such as organization and vocabulary but was weak on off-topic or nonsensical responses and inconsistent across student groups.
• Instructors. Pair AI writing tools with critical AI Literacy work on transparency, Privacy, and overreliance, keeping outputs transparent and editable so learners retain ownership of their self-regulated writing process — and treat AI literacy as an active design practice rather than a knowledge base, taking account of differential access across schools.

Limitations

• The synthesis rests on eight peer-reviewed studies (2019–2025) re-screened from a broader K-12 AI literacy review; a single study carries the writing-practice theme and three carry assessment, so no theme is more than weakly supported.
• The screening prioritized educational level but did not systematically account for students' writing proficiency, linguistic background (ESL/EFL or multilingual), or specific literacy development needs — factors that foundational writing research shows shape writing performance, and a gap the authors ask future work to close.
• Coding was done independently by two researchers with no formal inter-rater reliability statistic, so trustworthiness rests on documented audit trails and consensus discussion rather than a reported agreement coefficient.
• Several design recommendations lean on research outside the eight-study corpus, including an eye-tracking study of young learners' gaze patterns, so not every design claim is supported by the reviewed evidence itself; the authors call for longitudinal and design-based studies of teacher mediation, student–AI interaction, and the balance between human and AI-supported feedback in authentic classroom settings.

Citation

Amayou, J., Abedini, P., & Hutchison, A. (2026). Rethinking Elementary Education's Writing Instruction in The Age of Generative AI: A Systematic Review. EdArXiv preprint.