AI evidence extraction

Generative AI in Academic Publishing: A Bibliometric Analysis and Emerging Debates on Integrity, Authorship, and Language Equity

Ilone Paweloszek · JANOLI International Journal of Artificial Intelligence and its Applications · 2026

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

9/10

Relevance

Evidence

Citations

0.00

FWCI

This is an AI analysis. Read the peer-reviewed original at the publisher: https://doi.org/10.64758/5zcwqq98

Methodology & findings

Study design

Bibliometric analysis using Scopus database (2023-2025 publications) combined with thematic mapping.

Sample

N = 311, 1 group

Primary method

Descriptive statistics using Bibliometrix R package biblioAnalysis() function. Keyword frequency analysis with square-root scaling applied to balance dominant and less frequent terms. Co-occurrence network analysis using association strength normalization. Network visualization using Fruchterman-Reingold layout algorithm. No inferential statistical tests reported.

Main result

The bibliometric analysis of 311 Scopus-indexed documents revealed that "the keyword co-occurrence network reveals three major clusters: 1. Technological cluster (red) Centred on ChatGPT, artificial intelligence, large language model, and scientific writing, 2. Medical-ethical cluster (blue) including terms such as human, humans, medical education, controlled study, and natural language processing, 3. Publishing and integrity cluster (green) Comprising terms such as publishing, peer review, research ethics, authorship, editorial, publication, and plagiarism." The study found that ethical and integrity concerns are central rather than peripheral to the discourse on generative AI in academic publishing.

Reports effect sizes.

Research paradigm

Interpretive/mixed-methods (bibliometric analysis combined with qualitative thematic synthesis)

Author conclusions

"The academic community is no longer merely experimenting with AI-assisted writing-rather, it is actively negotiating the boundaries of legitimate, transparent, and equitable use of such tools. Across disciplines, a central tension emerges: generative AI can improve clarity, accessibility, and workflow efficiency, but it also poses risks of fabricated citations, obscured authorship, and diminished accountability. The discussion presented in this paper shows that neither extreme-complete prohibition nor unrestricted adoption-is sustainable. Instead, the evolving practice in leading journals and conferences points toward a middle-ground model that prioritises transparency, scientific responsibility, and linguistic fairness."

Risk of bias

Database selection bias: Scopus-only coverage may underrepresent discussions in specialized databases (PubMed, MEDLINE, Embase, ERIC, IEEE Xplore); Search strategy limitation: Single predefined keyword combination may miss related themes (AI-assisted peer review, automated workflows, discipline-specific frameworks); Metadata limitation: Authors' keywords reflect author terminology, not full conceptual content; themes like linguistic equity may appear indirectly; Temporal snapshot bias: 2023-2025 period represents exceptional technological/policy change; findings may not reflect stable research landscape; Temporal snapshot bias: 2023-2025 analysis captures rapidly evolving field; Publication language bias: Scopus predominantly indexes English-language publications; Database coverage bias: Scopus may differentially index biomedical vs. other disciplinary literatures; alternative formulations might retrieve different literature

Limitations

"The bibliometric analysis was based solely on Scopus-indexed publications
Scopus offers broad multidisciplinary coverage, but important discussions on generative AI-particularly those emerging in clinical, biomedical, and educational contexts-may be more extensively represented in specialised databases such as PubMed, MEDLINE, Embase, ERIC, or IEEE Xplore
As a result, the present study may underrepresent certain disciplinary nuances." Additionally, "the analysis relied on a single search strategy using predefined keyword combinations," and "the use of authors' keywords and metadata-although standard in bibliometric research-reflects the terminology selected by authors rather than the full conceptual content of the articles." Furthermore, "this study focused on publications from 2023 to 2025, a period of exceptionally rapid technological and policy change," making findings "a snapshot rather than a stable or mature research landscape."

Open questions raised

Need for broader database inclusion beyond Scopus (PubMed, MEDLINE, Embase, ERIC, IEEE Xplore) to capture disciplinary nuances
Alternative search strategies needed to capture AI-assisted peer review, automated editorial workflows, and discipline-specific ethical frameworks
More comprehensive examination of emerging themes (linguistic equity, conference submission policies, AI detection tools) that may appear indirectly in metadata
Longitudinal analysis beyond the 2023-2025 snapshot to understand how editorial guidelines and norms continue to evolve
Investigation of how generative AI transforms academic publishing across diverse disciplinary contexts
Full conceptual content analysis beyond author-selected keywords to better capture emerging themes (linguistic equity, conference submission policies, AI detection tools)

Data: Scopus database exportExtracted from: pdf

Explore related topics

Scholarly Publishing Scholarly Infrastructure Literature Review