Type: article
Abstract: Generative artificial intelligence is transforming research integrity in scholarly publishing by making misconduct easier to produce, harder to detect, and increasingly difficult to govern. We conducted a scoping thematic review of 68 peer-reviewed studies published from 2016 to 2026, analysed through inductive thematic analysis. Our review shows that, while prior studies document specific phenomena such as AI-generated text, hallucinated citations, paper mills, statistical anomalies, and post-publication scrutiny, the literature remains fragmented and largely descriptive. We identify five themes and fifteen sub-themes capturing the social, technical, and institutional dimensions of AI-mediated research integrity. The paper makes three contributions: it conceptualises AI-mediated research integrity as a sociotechnical governance problem; develops a constitutive mechanism-based model showing how interacting components produce and reproduce system-level integrity vulnerability; and identifies three persistent tensions: escalation between AI-enabled misconduct and AI-enabled detection, threshold ambiguity around acceptable AI use, and governance lag between technological change and institutional response.
Institution: The University of Queensland
Language: English
Subfield: Education
Field: Social Sciences
Domain: Social Sciences
Open Access status: green
OpenAlex ID: W7222912946