Dr Jesvir Kaur Mahil · September 2026
Academic integrity is often addressed only after something appears to have gone wrong: a high similarity score, unattributed material or uncertainty about whether a student’s work is genuinely their own. This article begins from a different ethical premise. Before treating problematic source use as misconduct, educators should ask what students have actually been taught about researching, paraphrasing, citation, authorship and responsible AI use. Drawing on an action-research study first undertaken in 2008–09 and revisited for the age of generative AI, I argue for an approach that combines clear expectations and accountability with teaching, practice, feedback and opportunities to learn. Academic integrity should not simply be policed; wherever reasonably possible, it should be taught.
Academic integrity is often discussed only when something has gone wrong.
A student submits work containing unattributed material, a similarity report appears concerning, or the authenticity of an assignment is questioned.
At that point, the institutional response can quickly become investigative and punitive.
My original study began from a different premise:
Before concluding that a student has intentionally cheated, educators should ask what the student has actually been taught about researching, paraphrasing, quotation, citation and the construction of an academic argument.
The original 2009 study distinguished deliberate unethical plagiarism from situations in which students were unfamiliar with referencing conventions, and argued for a supportive rather than punitive response where the problem represented a learning need.
That distinction is even more important now.
Students work in an information environment shaped by search engines, digital repositories, automated writing tools and generative AI.
They need explicit teaching about what counts as their own intellectual contribution, what must be attributed, when AI assistance is permitted, how it should be acknowledged, and how they remain responsible for the accuracy and integrity of the work they submit.
Fairness therefore requires more than enforcing academic-integrity rules. It requires ensuring that students have first been given a meaningful opportunity to understand and learn how to meet them.
The language used by educators matters.
Plagiarism describes a serious problem, but an educational approach begins with the broader positive goal of academic integrity: enabling students to participate responsibly in a community of knowledge.
Unintentional plagiarism can arise when students:
do not yet understand the difference between common knowledge, another person’s idea and another person’s exact words;
take notes in ways that blur the boundary between copied material and their own commentary;
paraphrase by changing a few words rather than reconstructing meaning in their own intellectual voice;
believe that adding a reference is sufficient even when wording has been copied too closely;
lack confidence in academic English or in the subject matter and therefore remain too dependent on source wording;
have experienced different educational conventions concerning memorisation, authority, collaboration or attribution;
misread a similarity percentage as a pass/fail plagiarism score;
or
use generative AI without understanding the institution’s rules about permission, acknowledgement, verification and authorship.
These possibilities do not remove the need to address deliberate misconduct.
They do, however, make diagnosis essential.
A fair response distinguishes a capability gap from intentional deception and uses teaching wherever teaching can reasonably correct the problem.
The original Supported Experiments project explored whether formative use of Turnitin could help students learn to use researched material more appropriately.
It asked two questions:
How can plagiarism in student assignments be reduced?
What is the impact of online similarity-detection software on student assignments?
The project focused on two Access to Higher Education classes, comprising 34 students. Their Business Context assignments required substantial research.
Students submitted work electronically, received feedback informed by Turnitin reports and, where appropriate, revised and resubmitted their work.
Ten students responded to an end-of-course survey. Approximately 70 teachers in the business division were invited to a staff survey and five responded.
The study was small-scale and practitioner-led. It was not designed to establish a causal effect or a universal similarity threshold.
Its value lies in the classroom process it documented:
students were shown evidence of matching text, given developmental feedback, taught how to improve their use of sources and allowed to demonstrate that learning in revised work.
The original report recorded a pattern of improvement over the year.
One student’s first assignment initially produced a 79% similarity result.
Across successive submissions, this fell to:
79% → 41% → 17%
Then, on the student’s first submission of the next assignment, the similarity result was:
5%
The original report interpreted this as evidence that formative feedback and repeated practice had helped the student change how researched material was used.
The student survey also pointed towards learning rather than mere deterrence.
Respondents described reading more widely, writing in their own words, summarising and referencing accurately. They also asked for better training in how to use the software.
Staff responses similarly suggested that similarity checking was most useful when embedded in a wider teaching and learning strategy rather than used in isolation.
An important outcome was procedural as well as academic.
In the previous year, I had dealt with three detected cases of plagiarism and other suspected cases that resulted in disciplinary consequences.
During the action-research year:
no students in the two classes faced disciplinary procedures.
Instead, they were supported to improve their research and writing skills.
The significance was therefore not simply that similarity results changed. Students were being given an opportunity to learn before mistakes became disciplinary problems.
The original report sometimes used a similarity percentage as shorthand for a “level of plagiarism.”
That terminology should no longer be used.
Turnitin’s current guidance is explicit that its Similarity Report identifies matching text; the percentage does not itself determine whether plagiarism has occurred.
Correctly quoted and referenced material can contribute to a match, while some forms of misconduct may produce a low similarity score.
Academic judgement and contextual review are therefore essential.
A fixed numerical threshold can encourage students to game a score rather than improve their scholarship.
The educational question is not simply:
“How low is the percentage?”
but:
“What is matching, why is it matching, and has the writer used and acknowledged sources appropriately?”
The similarity report becomes more valuable when students learn to interrogate individual matches and make reasoned revisions.
The 2009 study addressed copying from websites, books and other students.
In 2026, academic integrity also requires attention to AI-generated or AI-transformed text.
Contemporary approaches increasingly emphasise assessment reform, explicit communication about permitted AI use, AI literacy, authentic assessment, critical thinking and ethical reasoning rather than relying on detection alone.
The key principle is continuity rather than panic.
Students still need to understand the provenance of ideas and words, but they must now also be able to explain what role, if any, an AI system played in producing their work.
Institutional and assessment-specific rules vary, so educators should state them clearly rather than expecting students to infer them.
The original study suggests a practical alternative to treating academic integrity primarily as a matter of detection and punishment. In a formative approach, expectations are made explicit, students practise the required skills, difficulties are diagnosed, feedback is used for improvement, and students gradually take greater ownership of their academic work.
Educator: Explain authorship, citation, collaboration and AI rules before assessment begins.
Student learns: What counts as acceptable assistance and what must be acknowledged.
Educator: Use real extracts to demonstrate quotation, paraphrase, synthesis and referencing.
Student learns: How source material is transformed into an argument rather than merely reproduced.
Educator: Provide a low-stakes task using the same conventions required in the assessed work.
Student learns: How to apply the rules before marks or penalties are at stake.
Educator: Review source use and, where available, a similarity report with the student.
Student learns: How to distinguish legitimate matches from problematic borrowing.
Educator: Give specific developmental feedback and allow revision where policy permits.
Student learns: How to correct errors and improve future writing.
Educator: Ask the student to explain key ideas, sources, choices and, where relevant, AI use.
Student learns: How to take ownership of the intellectual process behind the submission.
Educator: Ask what changed between drafts and what the student will do differently next time.
Student learns: How academic integrity becomes a transferable writing habit.
Students who begin drafting while moving sentence-by-sentence through a source are particularly vulnerable to patchwriting.
Encourage them to read for understanding, close the source, make a brief note of the idea, and then explain it in their own conceptual language before returning to check accuracy.
Notes should clearly distinguish direct quotations, paraphrases, personal reflections and bibliographic details.
This simple habit reduces accidental loss of attribution and also helps students demonstrate their research process.
A sound paraphrase changes the construction of the explanation because the writer has understood and re-expressed the idea.
Merely substituting synonyms can preserve the structure of the original too closely.
Students should be taught to place sources in conversation: compare claims, identify differences, evaluate evidence and explain why a source matters to the writer’s own argument.
This moves the task from information gathering to knowledge construction.
Where students are allowed to see reports before final submission, teach them to inspect the matched passages rather than chase a target percentage.
Ask:
Is this a quotation?
Is the citation complete?
Is the wording too close?
Is the match a title, reference entry or standard phrase?
What revision would improve the intellectual ownership of the passage?
Where generative AI is permitted, students should be taught to follow the rules for the particular assessment, disclose or acknowledge use where required, verify factual claims and sources, protect confidential or personal information, and retain responsibility for the final work.
They should also understand that:
fluent AI output is not evidence of accuracy, scholarship or personal learning.
Academic integrity is strengthened when assessment makes learning visible.
In the generative-AI era, contemporary sector thinking recommends moving beyond reactive detection towards assessment design that promotes personal engagement, higher-order thinking, ethical reasoning and transparent use of technology.
Practical approaches include:
Use staged assessment: proposal, source map, outline, draft, feedback and final submission.
Ask students to connect theory to a specific case, experience, dataset, placement or local context.
Include short oral explanation, viva, presentation or reflective commentary where proportionate.
Ask students to submit an annotated bibliography or explain why particular sources were selected.
Require a brief process statement describing how the work developed and what tools were used.
Make AI permissions assessment-specific: prohibited, limited, permitted with acknowledgement, or deliberately integrated.
Assess evaluative judgement: the student’s ability to check, challenge and improve information rather than merely reproduce it.
These approaches should not be designed as traps.
Their educational purpose is to align assessment with the capabilities we actually want students to develop.
Before submitting an assignment, students can use the following questions to check whether their work demonstrates academic integrity and genuine intellectual ownership.
Then add each of these as ordinary Paragraph blocks. I recommend keeping each question bold so the checklist is easy to scan:
Can I explain the main argument in my own words without reading the assignment?
Have I clearly marked every direct quotation and cited its source?
Have I cited ideas, data, images or distinctive arguments that are not my own?
Have I paraphrased by reconstructing meaning rather than simply changing words?
Have I checked that every in-text citation has the required reference-list entry, and vice versa?
If I used a similarity report, have I reviewed the individual matches rather than focusing only on the percentage?
Do I know the rules for generative AI on this particular assessment?
If AI use is permitted, have I acknowledged or declared it in the way required?
Have I independently checked any factual claims, quotations and references suggested by AI?
Could I explain to my tutor how I researched, drafted and revised this work?
Academic integrity cannot be secured by policy documents and misconduct procedures alone.
If institutions want students to become responsible academic writers, they need to create repeated, contextualised opportunities to learn what responsible authorship means.
This requires educators to share a sufficiently consistent language around authorship, source use, referencing, collaboration and AI, while recognising that expectations may differ between disciplines and assessments.
Staff also need appropriate development so that similarity reports and emerging AI-related concerns are interpreted with academic judgement rather than through simplistic numerical thresholds or assumptions.
The central question therefore needs to shift.
Instead of asking only:
“How do we catch plagiarism?”
we should also ask:
“How do we design teaching, assessment and feedback so that students learn to become responsible authors?”
Detection still has a legitimate place. Deliberate academic misconduct must be addressed fairly and consistently.
But detection without education identifies a problem only after an important opportunity for learning may already have been missed.
A preventative approach places responsibility on both sides of the educational relationship: students are accountable for the integrity of the work they submit, while educators and institutions are responsible for making expectations transparent and providing meaningful opportunities to develop the capabilities required to meet them.
The original project was a small-scale practitioner action-research study, and its findings should be interpreted accordingly.
The study involved 34 students across two Access to Higher Education classes. Ten students responded to the end-of-course survey, while only five staff members responded to the staff survey.
There was no control group, no inferential statistical analysis and no independent qualitative coding of the responses.
I was also both the educator and the researcher. I entered the project with an existing concern that some apparent plagiarism represented gaps in students’ academic-writing knowledge rather than deliberate misconduct, creating the possibility of confirmation bias in how the findings were interpreted.
The study also predates generative AI by many years. Its original focus was principally on students’ use of material from websites, books and other sources rather than the much more complex questions of authorship, assistance and intellectual ownership now created by generative technologies.
There is one further limitation that becomes particularly important when revisiting the study in 2026.
The original report sometimes interpreted a declining similarity percentage as evidence of declining plagiarism. We should now make a more careful distinction:
a reduction in similarity is not, by itself, evidence that plagiarism has been reduced.
What the study more securely demonstrates is that students were able to review matched material, receive feedback, revise their writing and develop more appropriate practices for working with sources.
The observed changes between drafts, students’ reflections on what they had learned, and the absence of disciplinary cases within the two classes remain useful practice-based evidence for the value of formative academic-integrity education, while not establishing a universal causal relationship.
The central insight from this work is simple:
academic integrity is learned.
Students should not encounter expectations about referencing, paraphrasing, source evaluation, authorship and responsible use of technology only when they are suspected of having done something wrong.
These are academic capabilities that need to be explicitly taught, modelled, practised, discussed and refined through feedback.
A strong approach to academic integrity is therefore neither permissive nor primarily punitive.
It combines clear boundaries with meaningful education; accountability with support; and institutional expectations with genuine opportunities to develop the skills required to meet them.
Where deliberate misconduct occurs, institutions need fair and proportionate procedures for addressing it. But where problematic practice reflects a remediable learning need, education should be part of the response.
When I undertook the original study in 2008–09, the question was whether similarity-detection software could be used differently: could a technology associated with detecting plagiarism become a tool for learning?
In 2026, generative AI presents education with a related question.
Will these technologies lead primarily to greater surveillance and suspicion, or can their emergence prompt us to think more carefully about authorship, assessment, critical thinking, transparency and what it actually means for a student to learn?
The technologies have changed dramatically.
The educational task has not.
We still need to help students research critically, write authentically, acknowledge their intellectual debts, use available tools transparently and take responsibility for the claims they make.
Academic integrity is therefore not simply something institutions should demand from students at the point of submission.
It is something education should help students learn how to practise.
The following principles translate the argument of this paper into practical action for educators, programme teams and institutions.
1. Teach academic integrity as a positive capability from the beginning of a programme, rather than introducing it primarily through warnings about misconduct.
2. Use authentic disciplinary examples to teach quotation, paraphrasing, synthesis, referencing and intellectual ownership.
3. Provide opportunities for formative review, feedback and revision before high-stakes submission where assessment policy permits.
4. Teach students that similarity reports identify matching text that requires interpretation; avoid presenting a particular percentage as a universal measure of plagiarism.
5. Make expectations concerning generative AI explicit for each assessment, including whether its use is prohibited, limited, permitted with acknowledgement, or deliberately integrated into the task.
6. Develop critical thinking, source evaluation, evaluative judgement and verification as central components of academic integrity.
7. Design staged, contextualised and authentic assessments that make students’ thinking and learning processes more visible.
8. Distinguish fairly between deliberate academic misconduct and errors that indicate a remediable learning need.
9. Ensure educators receive appropriate development in both the pedagogy of academic integrity and the institution’s academic-misconduct procedures.
10. Evaluate academic-integrity initiatives longitudinally, looking beyond misconduct statistics to consider students’ learning, confidence, equity and progression.
Mahil, J. K. (2009). Using Formative Assessments to Reduce Plagiarism. Supported Experiments Project.
Race, P. (2007). The Lecturer’s Toolkit: A Practical Guide to Assessment, Learning and Teaching (3rd ed.). Routledge.
Keeley-Browne, L. (2007). Training to Teach in the Learning and Skills Sector. Pearson Education.
Petty, G. (2006). Evidence-Based Teaching: A Practical Approach. Nelson Thornes.
Turnitin. (2025–2026). Turnitin and plagiarism; Understanding the similarity score. Turnitin Guides.
Tertiary Education Quality and Standards Agency (TEQSA). (2023). Assessment reform for the age of artificial intelligence.
Tertiary Education Quality and Standards Agency (TEQSA). (2026). Academic Integrity Toolkit.
Tertiary Education Quality and Standards Agency (TEQSA). (2026). Assuring quality learning in a gen AI-integrated future: The role of adaptive capabilities.
Quality Assurance Agency for Higher Education (QAA). (2026). AI, assessment and a sector under pressure.
This article is a substantially revised and updated version of a Supported Experiments action-research project that I conducted in 2008–09, exploring the formative use of similarity-detection software with Access to Higher Education students.
The original classroom evidence has been retained as an important part of the paper’s history. However, the terminology, interpretation and educational implications have been reconsidered in light of developments in academic-integrity education, assessment practice and generative AI.
In particular, this revised article makes a clearer distinction between textual similarity and plagiarism and avoids treating a similarity percentage as a measure of academic misconduct.
The 2026 version also extends the original argument. Academic integrity is understood here not simply as compliance with rules, but as a set of capabilities that students can develop through explicit teaching, modelling, practice, feedback, critical judgement and reflection.
The emergence of generative AI makes this educational approach even more important. Questions of academic integrity now concern not only citation and source use, but also authorship, intellectual ownership, transparency, verification and responsibility for the knowledge claims we make.
The enduring principle connecting the original study with this updated paper is therefore:
Where a problem can reasonably be addressed through learning, education should be part of the response.
Dr Jesvir Kaur Mahil (September 2026)
© 2026 Dr Jesvir Kaur Mahil. All rights reserved.
This article may be shared by linking to this webpage. Please do not reproduce or republish the text without permission.