Authors : Syeda Akeefa; Dr. Girish Kumar D.; Sharvani V.; Jennifer Mary S.

Volume/Issue : Volume 11 - 2026, Issue 4 - April

Google Scholar : https://tinyurl.com/3nvvzxr3

Scribd : https://tinyurl.com/3tnstw4h

DOI : https://doi.org/10.38124/ijisrt/26apr2482

Note : A published paper may take 4-5 working days from the publication date to appear in PlumX Metrics, Semantic Scholar, and ResearchGate.

Abstract : The rapid growth of Large Language Models (LLMs) has transformed writing assistance and simultaneously enabled sophisticated forms of academic deception. Students now leverage AI paraphrasing tools that preserve semantic meaning while rewriting sentence structure, reducing lexical overlap, and masking direct borrowing from online sources. Traditional plagiarism detectors—designed primarily forever batim copying—struggle to detect these semantically equivalent but lexically divergent patterns. This paper presents a comprehensive, multi-layer AI-Powered Academic Integrity Assistant capable of detecting direct plagiarism, AI-assisted paraphrasing, semantic similarity obfuscation, and stylistic inconsistencies in academic writing.The system integrates vector-based retrieval, semantic drift scoring, paraphrase classification, textual entailment modeling, stylometric forensics, perplexity-based AI text detection, and a final-layer meta-classifier. The manuscript expands the system description to a full-length research format by providing in-depth architectural analysis, extensive methodology, detailed experimental insights, robust discussion of limitations, and an ethics-centered framework for deployment in academic institutions. The goal is to support universities in establishing reliable, transparent, and fair academic integrity monitoring infrastructures that evolve along side modern AI capabilities.

Keywords : Plagiarism Detection, AI Text Detection, Sty-Lometry, Semantic Similarity, Vector Retrieval, Large Language Models, Academic Integrity.

References :

1. OpenAI Research Group, “A technical overview discussing responsible deployment practices for modern text-generation models,” Open AI Publications, 2023.
2. S. Kirchenbauer and collaborators, “Ananalysis of statistical watermark-ing strategies for identifying AI-generated linguistic content,” Research Manuscript, 2023.
3. B.Zhang,Y.Li,andM.Chen,“Anempiricalsurveyandevaluationof neural paraphrase detection approaches,” IEEE Transactions onKnowledge and Data Engineering, 2022.
4. T. Mitchell, “Discussion of hybrid human–machine methodologies fortextualforensicsinacademicworkflows,”ArtificialIntelligenceReview,2 021.
5. T. Guo, S. Rao, and L. Wang, “Architectures and engineering practicesfor scalable deep retrieval and semantic search,” ACM Transactions onInformation Systems, 2021.
6. J.Stark,“ExploringethicalconsiderationsininstitutionaladoptionofAI-basedstudentevaluationtechnologies,”EducationandAIPolicyJournal,202 0.
7. M. Potthast and colleagues, “Insights and trends from the 2020 shared evaluation tasks on plagiarism and text reused etection,”CLEFWorkingNotes, 2020.
8. S.Corley,“Analysis of linguistic entropy signals for assessing authorship variation and writing irregularities,” Journal of Language and Information,2020.
9. M.McCarthy,“Researchonvariabilityinstylisticpatternsacrossdifferentform s of academic writing,” Academic Linguistics Review, 2020.
10. P. Juola, “An overview of stylometric techniques for determining authorship indigital text analysis,” Proceeding soft heLRECConference,2019.