CoChat

A Systematic Review of Generative AI in K–12: Mapping Goals, Activities, Roles, and Outcomes via the 3P Model

Open access
Lin, Xiaoling, Tan, Hao · 2025 · Systems

The DOI matches its Crossref record; 147 of the 147 references it cites were found in a registry or an index; 162 of the 314 sources its claims cite were checked for support.

Abstract

Generative AI is reshaping k–12 learning as a multi-agent system in which goals, activities, and roles co-evolve across formal and informal environments. Following PRISMA and appraising quality with MMAT, we synthesize 84 peer-reviewed empirical studies (2020–2025) involving learners aged 3–18. Using Biggs’s 3P model as a systems lens and embedding CIMO logic, we code learning objectives, activity designs, AI role paradigms, and outcomes. Seven recurring objectives emerge (language/literacy; STEM; creativity; socioemotional skills; feedback literacy and self-regulation; motivation; AI literacy). Five dominant activity patterns are identified: dialogic tutoring and formative feedback, generative iterative co-creation, project-based problem-solving, simulation/game-based learning, and assessment support. Across studies, AI roles shift from AI-directed to AI-supported/empowered, re-allocating agency among students, teachers, and caregivers via feedback loops. Reported outcomes span three categories—epistemic, practice, and affective/identity—with opportunities of deeper knowledge, improved practice, and stronger engagement, and risks of hallucinations, reduced originality, over-reliance, motivational loss, and ethical concerns. We propose a goal–activity–role alignment heuristic for instructional design, plus safeguards around teacher professional development, feedback literacy, and ethics. We call for longitudinal and cross-cultural research to evaluate the impacts of GenAI in k–12.

Overview

CoChat Summary

TL;DR

This PRISMA systematic review synthesizes 84 peer-reviewed empirical studies (2020–2025) on generative AI in K–12 (ages 3–18), using Biggs’s 3P model with CIMO logic to map goals, activities, AI roles, and outcomes. It identifies seven learning objectives, five activity patterns, and a shift from AI-directed to AI-supported/empowered roles, with outcomes categorized as epistemic, practice, and affective/identity.

Key findings

  • From 817 records (Scopus: 528; Web of Science: 289), 605 remained after de-duplication and removal of pre-2020 items; title/abstract screening excluded 470, leaving 119 for full-text review, and 84 studies with MMAT scores ≥3 were retained.
  • Seven recurring objectives emerged: language/literacy; STEM; creativity; socioemotional skills; feedback literacy and self-regulation; motivation; and AI literacy. Five dominant activity patterns were identified: dialogic tutoring and formative feedback, generative iterative co-creation, project-based problem-solving, simulation/game-based learning, and assessment support.
  • AI roles shift from AI-directed (learner-as-recipient) to AI-supported (learner-as-collaborator) and AI-empowered (learner-as-leader); reported outcomes span epistemic, practice, and affective/identity categories, with affective and identity outcomes most frequently reported.
  • Opportunities include deeper knowledge, improved practice, and stronger engagement; risks include hallucinations, reduced originality, over-reliance, motivational loss, and ethical concerns. The review also notes 26 studies targeting language and literacy enhancement.

Methods

  • PRISMA-guided systematic review searching Scopus and Web of Science between May and July 2025 for English-language, peer-reviewed journal or conference publications from 2020 onward; inclusion required empirical studies with 3–18-year-old learners and authentic teaching/learning activities using GenAI, while higher education, purely technical papers, attitude-only surveys, non-GenAI, and non-empirical work were excluded.
  • Quality was appraised with MMAT, retaining studies scoring ≥3; two reviewers independently screened all studies, and coding integrated CIMO-Logic with Biggs’s 3P model (Presage–Process–Product) across four research questions on goals, activities, roles, and outcomes.
  • Analysis coded learning objectives, activity designs, AI role paradigms, and outcomes; outcomes were reframed as Epistemic, Practice, and Affective/Identity.

Limitations

  • Stated: most included studies focused on short-term interventions, making sustained long-term impact difficult to assess; few studies systematically assessed algorithmic bias or privacy risks, potentially underestimating negative outcomes; and the review did not distinguish formal from informal learning environments or among age groups, which may mask contextual and developmental moderators.
  • Evident in the text: no pooled effect sizes or meta-analytic estimates are reported; the synthesis is thematic/narrative rather than quantitative.
  • Evident in the text: the search was restricted to English-language, peer-reviewed journal or conference publications indexed in Scopus and Web of Science, which may limit coverage of non-English or non-indexed work.

Model-written summary. Check it against the paper before relying on it.

Integrity signals

Registry

Crossref · 10.3390/systems13100840

Cites

147 of 147 references verified0 citation errors · 0 not checkable

Statistics

Nothing to recompute.

Verification

Claims the paper makes

72 Valid · 13 Not valid · 28 Not checked · 56 Own results and interpretations

46 more claims · 13 with their cited sources checked

Create a free account to see all 169 claims and to verify the 12 references still open.

See all claims — free account
  • It may state two separate claims as one. 104
  • Probably not a substantial claim. 30
  • Borderline: one of the tests for a substantial claim is unclear. 3

Create a free account to see these claims.

Claims other papers make about this one

Each is checked against this paper’s own text: does it say what the citing paper says it does?

Self-check

Percentages, sums and reported differences re-computed from the paper’s own numbers.

1 of 1

Cited works that were retracted

Works this paper cites that were later retracted, from the retraction notices publishers deposit. Major when the notice came out before this paper was published.

2 of 2

Citations

Cites

2 retracted

Research works printed but not linked: 10

  1. Dietz, G., Outa, J., Lowe, L., Landay, J.A., and Gweon, H. (2023, January 26–29). Theory of AI Mind: How adults and children reason about the “mental states” of conversational AI. Proceedings of the Annual Meeting of the Cognitive Science Society, Sydney, Australia.
  2. Salomon, G. (1997). Distributed Cognitions: Psychological and Educational Considerations, Cambridge University Press.
  3. Biggs, J., Tang, C., and Kennedy, G. (2022). Teaching for Quality Learning at University 5e, McGraw-Hill Education.
  4. Piaget, J. (1971). Biology and Knowledge: An Essay on the Relations Between Organic Regulations and Cognitive Processes, U. Chicago Press.
  5. Vygotsky, L.S. (1978). Mind in Society: The Development of Higher Psychological Processes, Harvard University Press.
  6. Street, B.V. (1984). Literacy in Theory and Practice, Cambridge University Press.
  7. Chen, 2024, A progressive prompt-based image-generative AI approach to promoting students’ achievement and perceptions in learning ancient Chinese poetry, Educ. Technol. Soc., 27, 284
  8. Chu, 2025, Effects of AI-generated drawing on students’ learning achievement and creativity in an ancient poetry course, Educ. Technol. Soc., 28, 295
  9. Li, 2025, A ChatGPT-AIPSS approach for SDGs, Educ. Technol. Soc., 28, 301
  10. Khuibut, W., Premthaisong, S., and Chaipidech, P. (2023, January 4–8). Integrating ChatGPT into synectics model to improve high school student’s creative writing skill. Proceedings of the International Conference on Computers in Education, Matsue, Japan.

Cited by

See all 22

Metadata

Authors
Lin, X., & Tan, H.
Affiliations
Hunan University (CN)
Corresponding author
Tan, Hao
Venue
Systems
Publisher
MDPI AG
Published
2025-09-25
Volume
13
Issue
10
Pages
840
ISSN
2079-8954 (ISSN-L)
OpenAlex
W4414511128
Cited by
22
References
134
Citations by year
2026: 22
FWCI
3.35
Type
journal article
Language
en
Open access
Locations
  • Systems (publishedVersion, cc-by)
  • DOAJ (DOAJ: Directory of Open Access Journals) (submittedVersion, cc-by-sa)
Topics
  • Social Sciences › Education › Artificial Intelligence in Education
  • Physical Sciences › Computer Science › Artificial Intelligence › Intelligent Tutoring Systems and Adaptive Learning
  • Health Sciences › Medicine › Health Informatics › Artificial Intelligence in Healthcare and Education
Keywords
generative AI in education, learning objectives, feedback literacy, AI literacy, student motivation, socioemotional skills, creativity, human-AI co-creation, game-based learning, assessment support, AI hallucination, teacher professional development, AI ethics, instructional design

About these checks

Registry, retraction, Cites and statistics results are computed, not judged, with one exception: a printed reference that no index lists is read once by a model before it counts as a citation error. Claims are extracted by a model and kept only when their quote appears verbatim in the text; each cited source is then checked, with a model’s help, for whether it says what the claim relies on. Those verdicts can be wrong — read the claim’s quote and the cited source before relying on one.

The claim verdicts this report counts are experimental: their accuracy is not yet measured.

A language model drew the claims from the paper on 2026-10-06, under version 3 of the claim definition (model deepseek-v4.1-flash, prompt extract-v3).

A second model labelled each claim on 2026-10-06 (model jev-1.13-20260917, definition revision gold-guideline-v0.2).

Cited sources are checked with a language model’s help (gpt-6-sol).

Statistics and arithmetic are recomputed, not judged.

Confidence bands are shown only for fields calibrated on our annotated test set. None is calibrated yet, so no band is shown.

What counts as a claim ›How we measure confidence ›How we check this ›

Are you an author, or do you see a problem with this page? Request a review

Data sources

Bibliographic data: Semantic Scholar, OpenAlex, Crossref.

Contains information from the Semantic Scholar Academic Graph Datasets, made available under the ODC Attribution License (ODC-BY 1.0).

Semantic Scholar’s data is described in: Kinney, R., et al. (2023). The Semantic Scholar Open Data Platform. arXiv:2301.10140.