AI Disclosure without Accountability: Paper Compliance and the Governance Limits of Transparency in Scientific Research

Abstract

This paper argues that institutionalising artificial intelligence (AI) disclosure in scientific research has led to a form of compliance that prioritises symbolic transparency over genuine accountability. Although journals and publishers are progressively mandating that authors disclose their use of AI, these policies remain fragmented, non-standardised, and largely unverifiable. Drawing on an exploratory review of 80 recent academic articles, the study identifies patterns suggesting that explicit AI disclosure remains limited and, when present, is primarily symbolic or narrative rather than verifiable. The paper explains these patterns by introducing the concept of paper compliance and the AI Disclosure Integrity Gap (AIDG), defined as the discrepancy between reported AI use and the actual epistemic influence of AI on research outputs. The analysis argues that this gap emerges from a misalignment between disclosure-oriented governance frameworks and the iterative, opaque, and often irreproducible characteristics of AI-assisted knowledge production. The study develops testable propositions and introduces the AI Use Traceability Framework (AUTF), a process-oriented governance model that emphasises traceability, auditability, and transparency. While recognising the practical, technical, and institutional constraints associated with implementing traceability mechanisms, the framework outlines proportionate pathways for strengthening accountability in AI-assisted research. The paper contributes to AI governance and research integrity by positioning disclosure as a necessary but insufficient mechanism and by highlighting the risk that current practices may create a false sense of transparency.