Plagiarism and AI Thresholds in Academic Theses: A Practical Quality Framework for Responsible Research
- 2 days ago
- 8 min read
Academic integrity is one of the most important foundations of higher education. In thesis writing, universities and academic institutions need clear standards to evaluate originality, proper citation, and responsible use of digital tools, including artificial intelligence. This article discusses a practical threshold model for reviewing plagiarism and AI-generated similarity in academic theses: less than 10% is acceptable, 10–15% needs evaluation, and above 15% is considered a fail. The aim is not only to detect problems but also to support students in producing ethical, transparent, and high-quality academic work. Using a simple academic approach, this article explains why thresholds matter, how they can be applied fairly, and how international and European universities often combine software results with human academic judgment. The article concludes that clear standards, student guidance, and responsible supervision can strengthen academic quality while encouraging honest research practices.
Introduction
Academic theses are important documents because they show a student’s ability to research, analyze, write, and contribute to knowledge. Whether at bachelor, master, or doctoral level, a thesis must reflect the student’s own understanding and effort. It should also respect the work of other scholars through correct citation and referencing.
In recent years, academic integrity has become more complex. Traditional plagiarism detection tools are now widely used by universities to check copied or closely repeated text. At the same time, artificial intelligence tools have created new questions about authorship, originality, and academic responsibility. AI can help students improve language, organize ideas, or understand difficult concepts. However, it can also create risks when students submit work that is not truly their own.
Because of this, institutions need clear and practical standards. A useful model is:
Less than 10% = Acceptable
10–15% = Needs Evaluation
Above 15% = Fail
This model gives universities, supervisors, and students a clear framework. It also supports fairness, because it does not treat every percentage in the same way. A low similarity score may be normal in academic writing because references, common terms, quotations, titles, and methodology phrases can appear in many papers. A medium score may need careful review. A high score may show serious problems and may require academic action.
For organizations such as the Euro-Arab Chamber of Commerce, which supports knowledge exchange, professional development, and cooperation between regions, academic integrity is also connected to trust. Strong research standards help build confidence between institutions, employers, learners, and society.
Literature Review
Academic integrity has been studied for many years as a key part of higher education quality. Scholars often define plagiarism as using another person’s words, ideas, structure, or work without proper acknowledgment. Plagiarism can be intentional, such as copying text without citation, or unintentional, such as poor paraphrasing or missing references.
Research on academic misconduct shows that students may plagiarize for different reasons. Some may lack writing skills. Others may not fully understand citation rules. International students may come from educational systems with different writing traditions. Some students may also face pressure, limited time, or weak supervision. Because of this, many universities now focus not only on punishment but also on education, prevention, and academic support.
European universities often promote academic integrity through a combination of policies, student handbooks, writing centers, thesis supervision, and plagiarism-checking systems. In many international universities, similarity reports are not treated as final judgments by themselves. Instead, they are used as evidence for academic review. A similarity score must be interpreted carefully because software can detect matching words, but it cannot always understand context, quality of citation, or whether the student has used sources correctly.
The rise of AI has added another layer to this discussion. AI-assisted writing tools can help with grammar, structure, translation, and clarity. Many universities are developing policies that allow limited and transparent use of AI, especially when it supports learning rather than replacing student work. However, submitting AI-generated content as original personal work can be considered a serious academic issue. This is why AI detection, like plagiarism detection, should be handled with care and combined with human judgment.
The literature also highlights the importance of proportionality. Not every case is the same. A thesis with 8% similarity may be acceptable if the matches are mainly references, technical terms, and properly quoted phrases. A thesis with 13% similarity may still be acceptable after review if the matches are correctly cited. However, a thesis with 20% or more similarity may raise strong concerns, especially if large sections are copied, poorly paraphrased, or not referenced.
Methodology
This article uses a qualitative policy-analysis approach. It reviews the practical meaning of plagiarism and AI thresholds in academic thesis evaluation. The article does not present statistical data from one institution. Instead, it develops a structured framework that can be used by universities, academic journals, chambers, training institutions, and quality assurance bodies.
The framework is based on three threshold levels:
Less than 10%: Acceptable
10–15%: Needs Evaluation
Above 15%: Fail
The analysis considers how these thresholds can be applied in real academic situations. It also discusses examples inspired by common practices in international and European higher education, where originality reports are often reviewed by supervisors, academic committees, or integrity officers.
The article focuses on four main questions:
What does each threshold mean?How should institutions review plagiarism and AI reports?How can students be supported before penalties are applied?How can universities maintain high standards while staying fair and educational?
Analysis
Less than 10%: Acceptable
A thesis with less than 10% similarity or AI-related concern is generally considered acceptable under this standard. This does not mean that the thesis is automatically perfect. It means that the detected level is low and normally within a reasonable academic range.
In many academic fields, some repeated language is expected. For example, standard research terms such as “the purpose of this study,” “the findings indicate,” or “the methodology is based on” may appear in many theses. References, titles, legal names, technical definitions, and commonly used theoretical phrases can also increase similarity scores.
A student writing a thesis in business, education, law, or management may use standard academic language. If sources are correctly cited and the student’s own analysis is clear, a score below 10% should usually not create concern.
This level supports a positive academic environment. Students should feel encouraged when their work shows strong originality. Supervisors can still provide feedback on citation style, paraphrasing, and structure, but the thesis can normally move forward for evaluation.
10–15%: Needs Evaluation
A thesis with a score between 10% and 15% should not be rejected automatically. Instead, it should be reviewed carefully. This is the “evaluation zone.” It means there may be areas that require attention, but the final decision should depend on academic judgment.
For example, a thesis may show 12% similarity because it includes a long literature review with many cited academic sources. If the references are correct and the student has properly paraphrased the material, the work may still be acceptable. However, if the same 12% includes uncited paragraphs, copied definitions, or repeated text from online sources, revision may be necessary.
European and international universities often use this kind of human review. A similarity report is not only a number; it is a map that shows where the matches appear. A supervisor or committee should check whether the matching text is in the reference list, direct quotations, methodology, appendices, or main analysis. The location of the similarity matters.
For AI-related concerns, the same balanced approach is needed. A tool may suggest that part of a thesis appears AI-generated, but AI detection is not always exact. The student may have used grammar correction, translation assistance, or writing support tools. Therefore, institutions should ask for clarification, review drafts, compare writing samples, and evaluate whether the student can explain the research.
At this level, the best response is usually educational. The student may be asked to revise, improve citations, rewrite weak sections, add personal analysis, or provide a declaration about AI use. This approach protects academic standards while helping students learn.
Above 15%: Fail
A thesis with a score above 15% should be considered a fail under this standard, especially when the similarity or AI concern appears in important parts of the thesis. This threshold sends a clear message that originality is essential.
A high score may show that the student has copied large sections, used poor paraphrasing, relied too heavily on sources, or submitted work that is not sufficiently original. If AI-generated content is detected above this level and the student did not declare or justify its use, the case may raise serious academic integrity concerns.
However, even when the standard says “fail,” institutions should still follow a fair process. The student should be informed of the issue, the evidence should be reviewed, and the decision should be documented. In some cases, the result may lead to resubmission, rewriting, academic warning, or referral to an academic integrity committee.
The purpose of this threshold is not to punish students unfairly. Its purpose is to protect the value of academic qualifications. A thesis must represent real learning, research ability, and independent thinking. If a thesis contains too much copied or machine-generated material, it cannot fairly represent the student’s achievement.
Findings
The first finding is that clear thresholds help reduce confusion. Students, supervisors, and committees benefit from knowing what is acceptable, what needs review, and what leads to failure.
The second finding is that numbers alone are not enough. A similarity or AI score should always be interpreted by qualified academic staff. A 14% score with correct citations may be less serious than a 9% score containing copied uncited paragraphs in the main findings.
The third finding is that academic integrity should be preventive, not only disciplinary. Students need training in citation, paraphrasing, research ethics, and responsible AI use. Universities that provide writing support often reduce misconduct because students understand expectations before submission.
The fourth finding is that AI policies must be transparent. Students should know whether AI tools are allowed, what kind of use is acceptable, and how to declare assistance. For example, using AI for grammar correction may be acceptable in some institutions, while using AI to generate full thesis sections may not be allowed.
The fifth finding is that international and European academic practice increasingly supports balanced evaluation. Many universities use software tools, but they also rely on supervisors and academic committees to make final decisions. This protects both academic quality and student fairness.
Conclusion
Plagiarism and AI thresholds are important tools for protecting academic quality in thesis writing. The standard of less than 10% acceptable, 10–15% needing evaluation, and above 15% fail provides a clear and practical framework for academic institutions.
This model supports fairness because it recognizes different levels of concern. It also supports quality because it sets a clear limit on unacceptable copying or undeclared AI-generated work. Most importantly, it encourages institutions to combine technology with human judgment.
Academic integrity should not be seen only as a rule or a punishment system. It is part of the learning process. Students should be guided to write honestly, cite correctly, think independently, and use digital tools responsibly. Universities and academic organizations can strengthen trust by applying clear standards, offering student support, and maintaining transparent review procedures.
For international education and Euro-Arab academic cooperation, such standards are especially valuable. They help create a shared culture of quality, responsibility, and respect for knowledge. In a global academic environment, originality is not only a technical requirement; it is a sign of professionalism, credibility, and intellectual honesty.
References
Bretag, T. (2016). Handbook of Academic Integrity. Springer.
Carroll, J. (2007). A Handbook for Deterring Plagiarism in Higher Education. Oxford Centre for Staff and Learning Development.
Howard, R. M., & Robillard, A. E. (2008). Pluralizing Plagiarism: Identities, Contexts, Pedagogies. Boynton/Cook Publishers.
Macdonald, R., & Carroll, J. (2006). Plagiarism: A complex issue requiring a holistic institutional approach. Assessment & Evaluation in Higher Education, 31(2), 233–245.
Pecorari, D. (2013). Teaching to Avoid Plagiarism: How to Promote Good Source Use. Open University Press.
Sutherland-Smith, W. (2008). Plagiarism, the Internet, and Student Learning: Improving Academic Integrity. Routledge.
Weber-Wulff, D. (2014). False Feathers: A Perspective on Academic Plagiarism. Springer.
Comments