Toward a New Approach for Examining Academic Citation Practices: Analyzing Citation Pattern Matches to Detect Structural and Paraphrased Plagiarism

The field of academic integrity is undergoing a profound transformation in the nature of scholarly plagiarism, particularly with the widespread use of paraphrasing tools and generative artificial intelligence. These technologies can now produce texts that differ linguistically from their sources while preserving the original conceptual structure. Traditional detection systems based on textual similarity, such as Turnitin and iThenticate, have shown clear limitations in identifying this emerging form of structural and paraphrased plagiarism, largely because they rely on lexical overlap rather than the ability to trace the intellectual framework of an argument. This review study aims to analyze the development trajectory of plagiarism detection tools over the past decade and to identify the methodological gaps that limit their ability to address structural plagiarism. It also evaluates the potential of the Citation Pattern Matching approach as an alternative and more reliable method for uncovering structural relationships between texts. The study relies on a systematic review of the literature from 2014 to 2024 using the PRISMA framework and applies precise criteria to select 33 studies distributed across three themes: text-matching tools, citation-based and bibliometric analysis, and semantic and AI-driven models. The results indicate that text-based tools perform well in detecting surface similarity but fail to uncover deep paraphrasing that preserves the original text’s referential structure and logical sequence. By contrast, the literature shows that citation analysis, including citation sequencing, repetition of core references, and network relationships among sources, represents a cognitive fingerprint that is difficult to manipulate and proves more effective in identifying structural plagiarism even when no clear linguistic similarity exists. The review also highlights the need for hybrid models that integrate textual, semantic, and citation-based approaches to develop a new generation of detection systems capable of addressing increasingly sophisticated forms of plagiarism. The study concludes that relying exclusively on lexical similarity ratios is no longer sufficient in the era of generative artificial intelligence, and that analyzing citation structures offers a more rigorous and reliable framework for assessing the originality of scientific texts. It proposes adopting Citation Pattern Analysis as a central step toward strengthening academic integrity and developing detection systems that keep pace with rapid transformations in knowledge production.