This paper presents AIRT (Anti-plagiarism Prompt for Integrity Research and Transparency) v4.1 — a structured AI-mediated protocol for detecting and auditing academic plagiarism, developed from the Front-End User (FEU) perspective. Unlike conventional BEU-centric (Back-End User) tools such as Turnitin or iThenticate that rely on database string-matching, AIRT leverages the full qualitative and quantitative capacity of Large Language Models (LLMs) as Probabilistic Meaning Mediators operating at 70–88% cognitive load. The framework introduces three integrated layers: Book Smart (academic standards A1–A4), Street Smart (FEU operational reality B1–B5), and Unifying Principles (C1–C4), culminating in a reproducible weighted scoring rubric (0–10 scale) and a structured audit report template. A critical revision in v4.1 separates AIGC declaration into two independent fields — Author Declaration (FEU/human) and Lingua's Impression (AI estimate from text patterns) — eliminating the paradox of AI self-reporting. The framework was field-tested against four papers submitted to UN Open Source Week 2026, yielding an average originality score of 8.25/10 with CLEAN status across all four. AIRT v4.1 is positioned as an open-source, accessible academic integrity tool for independent researchers operating within budget constraints, grounded in the philosophical principle of 自主者 (Zì Zhǔ Zhě): sovereignty of mind and integrity of thought.
Kian Tik Go (2026) studied this question.