31/08/2025
Hamaniuk, V.A., Semerikov, S.O. and Shramko, Y.V., 2025. Mapping the AI-plagiarism detection landscape: a systematic knowledge graph analysis of research evolution and critical gaps (2022-2025). Educational Technology Quarterly [Online]. Available from: https://doi.org/10.55056/etq.965
The emergence of ChatGPT in November 2022 triggered an unprecedented crisis in academic integrity, prompting a 550% surge in AI plagiarism detection research within a year. This study presents the first comprehensive knowledge graph analysis of this rapidly evolving field, systematically mapping 58 papers from the Scopus database (2022--2025) to understand research evolution, thematic clusters, collaboration patterns, and critical gaps. Using a novel three-phase methodology combining bibliometric analysis, content extraction, and interactive network visualisation, we constructed a knowledge graph revealing both the field's explosive growth and its fundamental fragmentation. Our temporal analysis identifies four distinct evolutionary phases - initial awareness (2022), reactive response (2023), method development (2024), and systematisation (2025) - yet this apparent maturation masks severe structural problems. While the field exhibits a unique balance between detection technology and educational approaches (41.4% each), this equilibrium proves conceptual rather than operational, with researchers working in extreme isolation (network density 0.0077, lowest documented for any emerging field). Critical gaps persist despite recognition: only 1.7% of papers address institutional policies, though 82.8% acknowledge their importance, zero longitudinal studies exist three years post-ChatGPT, and programming contexts remain severely underexplored (12.1% versus 25--30% expected). The collaboration network analysis reveals 44 disconnected components, with the largest containing only seven authors, indicating knowledge develops in silos rather than building cumulatively. Citation patterns show concerning concentration, with the top five papers capturing 51.8% of all citations, suggesting premature convergence around potentially limited perspectives. ChatGPT dominates research focus (43.1% of papers), creating vulnerabilities as the AI landscape diversifies, while Western institutions contribute 68.9% of research, limiting global perspective diversity. The interactive knowledge graph visualises these patterns, with gap nodes remaining peripheral to mainstream research networks, suggesting they will persist without intervention. These findings have profound implications: educators must shift from detection to AI literacy development, institutions need interdisciplinary policy task forces, not individual decision-makers, technology developers should pivot from the detection arms race toward educational enhancement tools, and policymakers must create adaptive frameworks for rapidly evolving capabilities. The study reveals a field in a pre-paradigmatic state, lacking shared frameworks, sustained research programs, or coordination mechanisms necessary for addressing complex challenges. We propose a transformation agenda that includes longitudinal cohort studies, international collaborative networks, participatory research with students, and research infrastructure investment. Ultimately, this analysis demonstrates that current fragmented, reactive approaches cannot address AI's comprehensive educational implications. The window for developing evidence-based responses narrows as AI capabilities advance. The educational community faces a choice: continue isolated efforts perpetuating gaps, or build collaborative infrastructure for transformation. The knowledge graph presented here provides both a warning and a roadmap, but navigating the future requires unprecedented coordination toward reconceptualising education for human-AI collaboration rather than defending against it.
