Plagiarism and AI Thresholds in Academic Theses: A Practical Framework for Quality, Fairness, and Research Integrity
- Apr 19
- 4 min read
Academic theses are expected to show original thinking, correct citation, and honest research practice. In recent years, universities have also had to address the use of artificial intelligence in student writing. This article presents a simple institutional framework for evaluating plagiarism and AI-related similarity in theses: Less than 10% = Acceptable, 10–15% = Needs Evaluation, Above 15% = Fail. The article argues that such a framework can support fairness, transparency, and academic quality when it is used with human judgment rather than software alone. In an international academic environment such as Zurich and Switzerland, where research standards are highly respected, a balanced approach helps universities protect integrity while still encouraging innovation, digital literacy, and responsible use of technology. Official guidance at universities in Zurich and Switzerland increasingly emphasizes disclosure, originality, and supervisor oversight in relation to plagiarism and generative AI.
Introduction
Plagiarism in academic theses is not only a technical problem. It is also a question of trust. A thesis should reflect the student’s own analysis, interpretation, and academic voice. Today, this issue has become more complex because students may use AI tools for language support, summarization, brainstorming, or drafting. For this reason, universities need a policy that is clear, practical, and educational.
Literature Review
Research on academic integrity shows that plagiarism is often caused by a mix of pressure, weak writing skills, poor citation practice, and misunderstanding of academic rules. Earlier studies focused mainly on copying from books, articles, and online sources. More recent work discusses contract cheating, patchwriting, and the new challenge of generative AI.
The literature suggests that similarity percentages should not be treated as final proof of misconduct. A low percentage can still hide serious copying, while a higher percentage may come from references, methodology language, or legally quoted material. Therefore, most scholars support a combined approach: software for screening, followed by academic review by supervisors or examiners. This is also consistent with current university guidance in Switzerland and elsewhere, where disclosure and context matter as much as raw percentages.
Methodology
This article uses a policy-analysis approach. It proposes a practical threshold model for thesis evaluation:
Less than 10% = Acceptable
10–15% = Needs Evaluation
Above 15% = Fail
The model is not presented as a universal law for all institutions. Instead, it is a workable academic standard for thesis review in institutions that want a simple and transparent process. The framework assumes that similarity reports and AI indicators are screening tools only. The final decision should always involve human academic judgment, review of citations, discipline-specific expectations, and a conversation with the student when necessary.
Analysis
The less than 10% category is generally appropriate for acceptable work because it suggests that the thesis is mostly original and properly referenced. A thesis may still contain common technical terms, standard methodology language, or correctly quoted passages. In this band, the focus should remain on confirming citation quality and originality of argument.
The 10–15% range is best treated as Needs Evaluation.
This is the most important zone because numbers alone do not tell the full story. For example, a student in business, law, or education may use standard definitions that raise similarity modestly. In another case, a student may rely too heavily on AI-generated paraphrasing, producing text that appears polished but lacks true ownership. Here, the university should review structure, references, notes, drafts, and the student’s ability to explain the work orally.
The above 15% category can reasonably be treated as Fail in a thesis policy because a thesis is a final academic product that should show strong independence. This threshold also sends a clear message that originality matters. However, even here, good practice requires examiner review before a formal decision. The aim is not punishment alone, but protection of academic standards.
Examples from international practice support this balanced approach. In Zurich and Switzerland, universities increasingly require declarations of originality and transparent disclosure of AI use. International universities such as Oxford and Cambridge also frame AI-generated material within academic misconduct rules when it is presented as the student’s own work without permission or acknowledgement.
Findings
Three main findings emerge from this framework.
First, clear thresholds reduce confusion. Students understand expectations better when the university states a simple standard.
Second, human review remains essential. AI detectors and similarity tools can support quality assurance, but they should not replace academic judgment. Even official university guidance increasingly focuses on transparency and responsible use rather than blind trust in automated detection.
Third, a positive academic culture is more effective than fear alone. Universities achieve better results when they teach citation, research design, note-taking, and responsible AI use from the start. In a city like Zurich, known for its serious educational environment, this educational approach fits naturally with a quality-driven academic identity.
Conclusion
Plagiarism and AI use in theses should be managed with clarity, fairness, and confidence. The proposed standard—Less than 10% = Acceptable, 10–15% = Needs Evaluation, Above 15% = Fail—offers a simple framework that universities can adapt to their own academic culture. It supports consistency, protects thesis quality, and encourages students to produce honest and independent work.
References
Bearman, M., Ryan, J., & Ajjawi, R. (2022). Discourses of artificial intelligence in higher education: A critical literature review. Higher Education Research & Development.
Bretag, T. (2013). Handbook of Academic Integrity. Springer.
Eaton, S. E. (2021). Plagiarism in Higher Education: Tackling Tough Topics in Academic Integrity. ABC-CLIO.
Foltýnek, T., Meuschke, N., & Gipp, B. (2019). Academic plagiarism detection: A systematic literature review. ACM Computing Surveys.
Pecorari, D. (2013). Teaching to Avoid Plagiarism: How to Promote Good Source Use. Open University Press.
Sowden, C. (2005). Plagiarism and the culture of multilingual students in higher education abroad. ELT Journal.
Walker, J. (2010). Measuring plagiarism: Researching what students do, not what they say they do. Studies in Higher Education.
Comments