AI evidence extraction

Cheating, Chatbots, and Change: A Bibliometric Mapping of Academic Integrity and Generative AI in Higher Education

Hanifah E. Daluma, Joel Aclao · Journal of Education and Learning Reviews · 2026

AI-generated evidence extraction, verified across multiple analytical personas. Not a substitute for the peer-reviewed original.

9/10

Relevance

1/4

Quality (LMQS)

Evidence

Citations

0.00

FWCI

This is an AI analysis. Read the peer-reviewed original at the publisher: https://doi.org/10.60027/jelr.2026.e2753

Methodology & findings

Study design

Bibliometric and science-mapping design using VOSviewer software.

Sample

N = 303, 8 groups

Primary method

Descriptive bibliometric techniques (annual output, publication distribution by country/source, author productivity, citation performance). Network-based science-mapping using VOSviewer software: co-authorship networks (minimum 2 documents per author), keyword co-occurrence networks (minimum 5 occurrences per keyword, with thesaurus-based harmonization), source networks (minimum 5 documents per source), and citation-based analysis (documents with ≥20 scholarly citations). Full counting used for co-authorship and keyword co-occurrence. Spreadsheet functions used for descriptive indicator computation.

Main result

The study found that publication output increased substantially from 56 publications in 2023 to 138 in 2025, with "output increased from 56 publications in 2023 (18.5%) to 108 in 2024 (35.6%) and 138 in 2025 (45.5%)". Additionally, "the field's initial agenda has been shaped by synthesis and guidance-oriented scholarship, followed by more specialized empirical work," and the Netherlands contributed the most publications (51), followed by the United Kingdom (33) and the United States (29).

Reports effect sizes.

Research paradigm

Positivist/Quantitative (bibliometric and science-mapping)

Author conclusions

"Overall, the study provides a baseline view of how integrity-focused GenAI scholarship organized itself in the first three years after ChatGPT's release: rapid growth, clear thematic branching, and early influence concentrated in a small set of venues and publications, interpreted with appropriate caution due to missing affiliation metadata and the intentionally bounded query."

Risk of bias

Selection bias from Boolean search string explicitly including misconduct-related terms (plagiarism, cheating, misconduct), making the corpus a misconduct-salient subset rather than full GenAI-in-higher-education literature; Missing affiliation metadata for 55.1% of records (167 of 303); Language restriction to English-only publications; Time window bias: 2025 publications had less time to accumulate citations at data extraction; First-mover advantage bias in citation counts for 2023 publications; Selection bias: Search string explicitly incorporated misconduct-related terms (plagiarism, cheating, misconduct), creating a misconduct-salient rather than comprehensive subset; Missing affiliation metadata: 55.1% of records (167/303) lacked country metadata, limiting geographic analysis; Temporal bias: 2025 publications had minimal time to accumulate citations, affecting citation-based comparisons; Publication venue bias: Analysis limited to English-language publications only; Database bias: Data from single source (The Lens.org) only, not cross-normalized with other bibliometric databases; Selection bias: Search string explicitly includes misconduct-related terms (plagiarism, cheating, misconduct), making the corpus a misconduct-salient subset rather than full GenAI-in-higher-education literature; Completeness bias: 55.1% of records lacked country affiliation metadata, limiting geographic analysis; Language bias: Only English-language publications retained; Citation window bias: 2025 publications had less time to accumulate citations at time of data extraction, affecting citation-based comparisons across years; Database bias: Data limited to The Lens scholarly works collection; other databases may contain additional relevant publications; First-mover advantage: Highly cited publications concentrated in 2023, reflecting early-entry effect rather than cumulative research value

Limitations

The authors acknowledge that "because the search string explicitly incorporated terms such as plagiarism, cheating, and misconduct, the resulting dataset is oriented toward studies that foreground breaches of academic integrity and institutional responses", noting that "this emphasis is treated as a deliberate focus of the study and later acknowledged as a limitation in relation to more capability-oriented GenAI research." Additionally, "author affiliation country information was available for 136 records
167 records (55.1%) lacked country metadata", and the authors note that findings should be "interpreted with appropriate caution due to missing affiliation metadata and the intentionally bounded query."

Open questions raised

The authors acknowledge that the field has been "shaped more by agenda-setting syntheses and policy framings than by mature, cumulative empirical programs." They note that "the most visible early contributions are largely reviews and framing papers, alongside work on assessment and detection," suggesting gaps in empirical research beyond synthesis and guidance-oriented work.
The authors implicitly identify a research gap regarding "more capability-oriented GenAI research" as opposed to the breach-focused literature mapped in this study. The dominance of "synthesis and guidance-oriented scholarship" over "mature, cumulative empirical programs" suggests a need for more sustained, empirical research programs beyond early reviews and policy framings.

Data: Data retrieved from The Lens (Lens.org) Scholarly Works collection on October 30, 2025. Final dataset comprised 303 publications on academic integrity and generative AI in higher education published between 2023-2025. Exported in CSV format including titles, authors, source information, abstracts, keywords, and citation counts.Extracted from: pdfAgreement 66%

Explore related topics

Citation Integrity Research Integrity Meta-Analysis