CoChat
Sign Up
Cited by 21Artificial Intelligence in EducationArtificial Intelligence in Healthcare and EducationIntelligent Tutoring Systems and Adaptive Learning

ChatGPT Interventions in Higher Education: A Systematic Review of Experimental Studies

Jin, Yiwen, Sercu, Lies·2025·Journal of Computer Assisted Learning
DOI 10.1111/jcal.70072
Open this paper in CoChat — free account

Abstract

Background Artificial intelligence has been reshaping many industries, and education is no exception. When ChatGPT burst onto the scene in late 2022, it ignited conversations among educators and researchers alike. Despite growing interest and experimentation with this technology in university classrooms, the field has been missing a comprehensive analysis of the actual research examining how ChatGPT is being used in higher education settings. Objectives This review addresses that gap by focusing on experimental designs and interventions that incorporate ChatGPT in higher education settings, with particular attention to its impact on academic outcomes. Methods This study followed the PRISMA guidelines to identify and analyse relevant experimental studies. Through systematic coding and analysis of 21 selected studies, this review examined research design, intervention types and academic outcomes. Results and Conclusions This review found that research on ChatGPT‐based interventions in higher education has focused on language and STEM domains. Although many of the studies are true experimental designs, they often lack large, diverse samples. Two main types of interventions are identified: task assistance and general learning support, each with unique strengths and challenges. ChatGPT was generally effective for knowledge acquisition, but its impact on skills development varied. Medium‐term interventions showed the best results, while short‐term effects were limited and long‐term…

Integrity signals

Registry

Crossref · 10.1111/jcal.70072

Cites

18 of 18 references verified0 citation errors · 0 not checkable

Statistics

Recomputing p-values needs the paper’s full text.

Claims the paper makes

This paper’s full text isn’t openly available, so its claims and self-check weren’t run. The registry and citation checks still apply.

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?

  • Jin and Sercu (2025) analyzed 21 experimental intervention studies and found that effects on skills varied by intervention type and duration.

    “Jin and Sercu (2025) analyzed 21 experimental intervention studies and found that effects on skills varied by intervention type and duration.”

    From From Assistance to Substitution: Generative AI and Undergraduate Research Skills (2026) · cites this paper as Jin and Sercu, 2025

  • Jin and Sercu (2025) show that intervention design and duration matter.

    “Jin and Sercu (2025) show that intervention design and duration matter”

    From From Assistance to Substitution: Generative AI and Undergraduate Research Skills (2026) · cites this paper as Jin and Sercu (2025)

  • Jin and Sercu (2025), reviewing 21 experimental studies, found that ChatGPT was generally effective for knowledge acquisition but that effects on skills varied.

    “Jin and Sercu (2025), reviewing 21 experimental studies, similarly found that ChatGPT was generally effective for knowledge acquisition but that effects on skills varied.”

    From From Assistance to Substitution: Generative AI and Undergraduate Research Skills (2026) · cites this paper as Jin and Sercu, 2025

  • Medium-term interventions appeared more promising than short-term interventions, while long-term outcomes were mixed.

    “Medium-term interventions appeared more promising than short-term interventions, while long-term outcomes were mixed.”

    From From Assistance to Substitution: Generative AI and Undergraduate Research Skills (2026) · cites this paper as Jin and Sercu, 2025

  • The distinction between assistive and substitutive use helps reconcile evidence showing that GenAI can improve the performance of academic students while also raising concerns about independent thinking, source evaluation, and long-term skill development.

    “This distinction helps reconcile evidence showing that GenAI can improve the performance of the academic students while also raising concerns about independent thinking, source evaluation, and longterm skill development (Deng et al., 2025; Gu & Yan, 2025; Jin & Sercu, 2025).”

    From From Assistance to Substitution: Generative AI and Undergraduate Research Skills (2026) · cites this paper as Jin & Sercu, 2025

Cites

Flagged first, then by citations

Cited by

Sorted by citations

Authors
Jin, Y., & Sercu, L.
Affiliations
KU Leuven (BE)
Corresponding author
Sercu, Lies
ORCID
Venue
Journal of Computer Assisted Learning
Publisher
Wiley
Published
2025-06-05
Volume
41
Issue
4
ISSN
  • 0266-4909 (ISSN-L)
  • 1365-2729
DOI
10.1111/jcal.70072
OpenAlex
W4411072120
Cited by
21
References
18
Citations by year
2026: 14 · 2025: 7
FWCI
3.22
Type
journal article
Language
en
Open access
closed
Licence
onlinelibrary.wiley.com/termsAndConditions#vor (vor)
Locations
  • Journal of Computer Assisted Learning (publishedVersion)
  • Lirias (KU Leuven) (submittedVersion)
Topics
  • Physical Sciences › Computer Science › Information Systems › Artificial Intelligence in Education
  • Health Sciences › Medicine › Health Informatics › Artificial Intelligence in Healthcare and Education
  • Physical Sciences › Computer Science › Artificial Intelligence › Intelligent Tutoring Systems and Adaptive Learning
Keywords
ChatGPT, large language models, higher education, AI-based interventions, educational outcomes, task assistance, learning support, knowledge acquisition, skill development, experimental design, language learning, STEM education

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.

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