AI Governance in Scholarly Publishing: A Computational Analysis of Policy Structures and Gaps

Abstract

The study examines the emerging landscape of artificial intelligence (AI) tools usage guidelines in scholarly publishing. It focuses on structure and gaps in existing policy frameworks through the analysis of twenty policy documents from academic publishers, research societies, open-access platforms, and universities. An integrated computational text analysis approach is used, including topic modeling, sentiment analysis, TF-IDF lexical analysis, and co-occurrence network analysis. The results indicate a pattern of structural convergence across publishers; however, this convergence is superficial and not seen in enforcement-oriented elements such as prohibition sections. The policy discourse is predominantly found in a neutral-positive tone. The discourse is conceptually rooted in legal and ethical terms, including authorship, privacy, and copyright. The thematic integration is limited, and a fragmented policy design is reflected in the documents. The study lays emphasis on data-driven, computational mapping of AI tools usage guidelines and policies in scholarly publishing. The current approaches reflect gradual institutional adaptation rather than coherent regulatory design.