AI-ASSISTED REWRITING AND IDEA LAUNDERING IN ACADEMIA: A PROOF-OF-CONCEPT STUDY FROM THE PERSPECTIVE OF PUBLICATION CONTROLS

Abstract

This study examines the extent to which the similarity reports and AI writing reports used in academic publishing can provide assurance against the risk of “idea laundering” that may be carried out through AI-assisted rewriting. Its central claim is that AI-assisted writing does not, in itself, constitute plagiarism; the ethical problem lies in rewriting another author’s academic idea, argument, or contribution without attribution and presenting it as an original contribution. To this end, the discussion/conclusion sections of twelve English-language social science articles in the fields of internal audit, corporate governance, and risk management were used. Three separate files were prepared for each source text: the original text taken from the source article, a naive LLM rewriting produced without any specific instruction, and a structural rewriting created to observe the risk of idea laundering under controlled conditions. This yielded thirty-six core evaluation files for the twelve source articles, together with five AI-assisted control texts grounded in the authors’ own original ideas. All files were uploaded to Turnitin in “No Repository” mode. The findings show that the original texts were recognized by Turnitin with an average similarity of 99.75%, that the average similarity score fell to 44.50% under the naive rewriting condition, and that it reached 19.08% under the structural rewriting condition. Two independent expert raters found that the main idea, argument flow, and concluding claim were largely preserved across all twelve cases. While the control texts returned a similarity score of 0%, their high AI writing scores support the position that an AI score cannot be interpreted as an indicator of plagiarism. The study argues that countering the risk of idea laundering requires a multi-layered publication governance approach grounded in semantic evaluation, source–argument traceability, editorial judgment, and an internal control perspective.