AI evidence extraction

Aplicaciones de inteligencia artificial en la escritura y la corrección académicas en la universidad: una revisión sistemática

Adriana Pérez, Steven K. McClain, Alana Roa Narváez, Nayibe Rosado, Lina Trigos-Carrillo, Heydy Robles et al. · Íkala Revista de Lenguaje y Cultura · 2025

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

2.66

FWCI

Top 10%

Impact

This is an AI analysis. Read the peer-reviewed original at the publisher: https://doi.org/10.17533/udea.ikala.355878

Methodology & findings

Study design

Systematic review following PRISMA 2020 Declaration guidelines.

Main result

The systematic review found that "AI improves linguistic quality and feedback in the writing process" and that "a notable increase in publications related to AI and academic writing was observed between 2022 and 2024, with the United States, China, Australia, and Canada leading in this area." The findings also indicate that "AI can expedite the writing process, foster learner autonomy, and enhance idea generation and refinement" while suffering from "shortcomings related to privacy risks, subtle error detection, and handling nuanced conceptual material."

Research paradigm

Critical realist; interpretivist (synthesizing qualitative and quantitative evidence on AI applications in academic writing)

Author conclusions

The authors conclude: "Although current literature suggests that incorporating ai-based tools into academic writing and editing may yield meaningful improvements in student learning outcomes, the body of research still needs to be expanded. Future studies should include a broader spectrum of discourse communities, academic genres (extending beyond scientific articles to books, book chapters, and novels), and diverse educational contexts and languages. Such investigations will help guide the equitable, responsible, and context-sensitive implementation of ai technologies, ensuring that these tools serve as valuable complements to, rather than substitutes for, human expertise and traditional pedagogical practices."

Risk of bias

Selection bias: Limited to open-access articles in two databases (Scopus, WOS); excludes research in other databases and languages beyond English/Spanish; Language bias: Only English and Spanish articles included; Publication bias: Only journal articles included; excludes grey literature, conference proceedings, dissertations; Geographic bias: Dominated by US, China, Australia, Canada publications; Recency bias: Focus on 2019-2024 publications may overrepresent recent trends; Quality heterogeneity: Mix of exploratory, descriptive, quantitative, qualitative, and mixed-methods studies without stratification; Selection bias: Limited to Web of Science and Scopus only; excluded non-indexed publications; Access bias: 714 documents excluded due to limited accessibility/non-open access; Publication bias: Recent publications (2023-2024) have not yet gained significant citation visibility; Methodological heterogeneity: Studies varied in scope and rigor (45.7% qualitative, 29.3% quantitative, 22.8% mixed-methods); Geographic bias: Overrepresentation of United States, China, Australia, Canada studies; Selection bias: Limited to open access articles (714 documents excluded for access restrictions); Database bias: Limited to Scopus and Web of Science only; Publication bias: Recent articles (2023-2024) have not yet gained significant visibility/citations; Geographic bias: Study distribution concentrated in USA, China, Australia, and Canada; Study design bias: Majority of included studies were exploratory and descriptive (45.7% qualitative, 29.3% quantitative, 22.8% mixed methods)

Limitations

"The analyzed literature primarily comprises exploratory and initial studies, often limited to small, case studies or specific text corpora
This represents a limitation in the understanding of the implementation of AI-based technologies in the correction, editing, and composition of academic texts written by college students." Additionally, "this SR limited its search to Scopus and WOS databases that, although large and international in scope, could exclude other published research
Furthermore, this study only included scientific articles published in English or Spanish." The authors note that "714 documents were excluded from the review due to limited access" which "can be considered a methodological limitation."

Open questions raised

Need for broader spectrum of discourse communities and academic genres beyond scientific articles
Research needed in diverse educational contexts and languages (beyond English/Spanish)
Lack of effective pedagogical strategies for classroom implementation of AI tools
Limited understanding of AI's capacity to handle complex rhetorical and conceptual challenges
Need for studies on data privacy, academic integrity, and equitable access
Underexplored connections between AI tools and university instruction/traditional pedagogy

Data: 92 articles included in the review with appendices available (Appendix 1 lists all included articles; Appendix 2 contains matrix of impact factors, quartiles, and citations with link: https://tinyurl.com/2a73ryoc)Extracted from: pdfAgreement 74%

Explore related topics

AI Manuscript Writing Academic Writing Literature Review