Tools facilitate cheating, or partner supports learning? GenAI and academic integrity issues from students' perspectives

Synthesis: Mulisa and Mezgebu interviewed 27 undergraduates at Ambo University in Ethiopia, in Afan Oromo, about how they understand Academic Integrity when they use GenAI for coursework. The students' accounts are divided against themselves: almost all use or watch peers use these tools, most credit them with improving their academic achievement, yet many also describe an uneven playing field in which AI-assisted work earns better grades than honest effort, and a small but articulate minority call GenAI use outright misconduct. The authors read this as an integrity-utility trade-off and argue that students' beliefs, not only institutional rules, drive ethical use, so awareness, clear policy, and Assessment redesign must arrive together.

Key Findings

1. Use is near-universal, and the benefits students name are concrete. Almost all participants reported using GenAI, or having friends who do, to create coursework and research projects. They credited it with raising productivity, refining their thinking, clarifying their language, and letting them explore ideas immediately with less effort, plus overcoming language barriers, finding up-to-date resources, saving time, and easing library searches.
2. Most students see the tool as an academic enhancer, not a threat. The majority believed GenAI enhanced their academic achievement rather than compromising integrity, while a small number treated its use for coursework as serious academic misconduct and two said they needed a clear legal framework before judging whether it is misconduct at all. The authors contrast this with Lund et al. (2025), where most students called such use severe misconduct, and attribute the difference to exposure to policy and levels of awareness.
3. Plagiarism is contested on technical grounds. Students who denied plagiarism reasoned that they copy nothing: "I do not copy and paste any intellectual property of other authors... All the ideas are my own; the apps make them more meaningful, coherent, and flawless. So where do I plagiarize?" (P6). Others reversed this, tying integrity to the purpose of education: "I think it should be absolutely criminal to use tools like Gemini and ChatGPT for coursework." The same participant (P23) admitted the practice: "Many of my friends, myself included, especially when we are running out of time, copy entire ideas directly from the AI, print them out, and submit them to our instructors. But we do not know what was done, either partly or at all."
4. Originality is where the confusion is deepest. Several participants claimed authorship because there were no other authors ("If it is not my original idea, then whose?", P2; "I did not cheat anybody; I got help that nobody ever gave me," P26), some were unsure whether to call their work original, and four concluded it did not represent them. Answering the authors' challenge that AI merely remixes existing online material, one participant moved to a position the authors record as new for them: "even if it was my idea and I did not compose it, someone else did it, a co-author. Therefore, AI should be recognized as a co-author of the idea. But AI is not a person; how can it be a co-author with a human being? Still, I think we need a solution that transcends us" (P17).
5. Fairness of assessment is the sharpest complaint. Almost all participants held that GenAI creates an uneven playing field between those who use it and those who do not: academic performance deviates from real ability where AI checks are not run, independent workers score lower while AI users score higher, and the problem extends into classroom testing where digital tools are not banned. Students framed this as a reward for less effort ("It is getting smarter to use AI for coursework and earn better grades, unless it is not recognized," P21) and as a demotivating injury to diligence: "They earn better grades than I, who worked hard without knowing what others have done... This has come to kill my sense of diligence in my academic pursuits" (P14). One participant (P19) summed up the uncertainty: "we, as students, are not sure whether we are benefiting or suffering from the use of AI."
6. Competition pulls students in, and they intend to stay. Students reported being pushed toward the tools to keep pace with stronger peers ("AI makes it easier for me to compete with stronger students. I am taking advantage of the opportunity," P4) and to avoid losing an assignment when time runs short (P9: "I would rather use AI to take what comes, if it is risky too, than leave my assignment behind"). A considerable number said they would continue using GenAI if nothing prevented them.

How the study was conducted

The study is Qualitative Research in design, using reflexive thematic analysis to capture how participants make sense of their lived experience of GenAI and integrity in higher education. Participants were selected by typical case sampling for relevant experience: 27 undergraduates from different fields of study at Ambo University in Ethiopia, 18 male and 9 female, with sample size settled by data saturation. Data were collected between November 2025 and January 2026 through individual semi-structured interviews lasting 38 to 47 minutes, audio-recorded against a prepared protocol covering familiarity with tools, coursework use, peers' use, benefits and risks, plagiarism, authorship, and assessment impartiality.

Interviews were conducted in Afan Oromo, the participants' native tongue, transcribed verbatim, translated into English and back-translated with AI assistance, then checked by a linguistic expert; the English transcription was coded and reported. Analysis moved through seven documented phases, from immersion and initial noting through pattern identification, merging into main and sub-themes, case-sensitive reflection, cross-participant comparison, and refinement. For trustworthiness the researchers debriefed after each interview and shared the analysis back with available participants, around half of whom contributed to confirmability. Ethical approval came from the Institutional Review Board of the Ambo University Institute of Education and Behavioral Sciences, with oral informed consent; all participants were over 18, and their contributions are reported under codes P1 to P27. The authors note that self-reported interview data expose the findings to social desirability bias.

Plagiarism, originality, and authorship

The three thematic areas map directly onto the study's research questions, and the authors describe the combined picture as paradoxical. On plagiarism, the paper's diagnosis is that students hold a technically defensible but incomplete definition: if plagiarism means reproducing someone else's words, a machine-written text that copies nothing is not plagiarism. The authors argue this reading misses what misuse costs the student, and they take up Ka and Chan's (2025) term "AI-plagiarism" for the case where a student submits machine-generated ideas as their own, following Laflamme and Bruneault (2025) in holding that the definition of plagiarism must expand beyond copying and pasting. Their stated worry is that a student who outsources the work loses the opportunity to be a critical and creative thinker on that topic.

On originality, the paper defines the construct as the student's own unique critical thinking, creativity, and evaluation of existing ideas, and concludes that AI-generated work should not count as original student work, because authorship involves more than remixing existing ideas. The authors call idea generation by prompt digital ghostwriting and tie it to a wider concern about dependence: if the student's role reduces to entering keywords, cognitive functions decline and students become dependent rather than active knowledge creators, which the paper links to technological cognitive atrophy.

Implications for policy and assessment

The paper's practical argument is that a purely punitive response is less promising than a balanced one. It recommends clear, comprehensive, and well-communicated institutional GenAI policies, targeted awareness raising about plagiarism, originality, and assessment inequity, and promotion of digital literacy and AI literacy. They note that where policies existed, students found them vague and faculty lacked the training to enforce them. Because preparing a policy is not the same as enforcing it, the authors call for reinforcement of rules and continuous evaluation of whether they work, and identify as a future research question whether policy or ethical awareness better mitigates GenAI risk, since studies of policy effectiveness remain scarce.

Assessment reform is the second lever. The authors urge a shift from intensive paperwork toward progressive learning processes, authentic assessments, reflective debriefings, and real-life project work, and toward tasks that assess higher-order thinking rather than the reproduction of text. They also point out the limits of the detection route: AI-generated work is not reliably detectable, and detectors return false positives and false negatives, so a stronger signal for detection-based integrity enforcement is not available. Underneath both levers sits the paper's framing of students' relationship to the tools: the closing line asks for GenAI to become a partner that supports learning rather than an instrument of academic misconduct, and the students' own testimony shows why the distinction is not settled: the same tool is experienced as an instant, always-available mentor and as an unfair advantage that rewards the least diligent users.

Limitations and open questions

The authors state four limits. The study could not capture the perspectives of students with disabilities, so those views are absent. The sample was mainly male because fewer female students were willing to take part, which may have limited representation of women's views. It was conducted in a single higher education institution, so transferability is confined to similar institutional and contextual settings. And the data are self-reported, leaving the findings open to social desirability bias. What the study does provide is a documented case of an integrity-utility trade-off from inside a student body, plus the authors' own claim that students' beliefs predict their behaviour more strongly than institutional rules do (Ka and Chan 2025).

Connected Concepts

• Academic Integrity — the paper's central construct, framed through honesty, trust, fairness, respect, and responsibility
• Learner Identity — students' claims to authorship when a machine composed the prose
• Equity — the uneven playing field between AI users and independent workers
• AI Misuse and Learning Harm — loss of critical thinking, creativity, and authorship through outsourcing
• Reducing AI Misuse — policy, awareness, and assessment redesign as mitigation
• Authentic Assessment — the paper's proposed replacement for intensive paperwork
• Assessment Validity — assessment results that no longer track students' real ability
• Motivation — demotivation of diligent students as a reported consequence
• Cognitive Offloading — "technological cognitive atrophy" and growing dependence on chatbots
• Framing AI Use for Students — whether GenAI reads as a cheating box or a calculator-like support
• Educational AI Policy — vague policies, weak enforcement, and unevaluated rules
• Global South — an Ethiopian university context for the integrity debate
• Multilingual Learning — language barriers named as the primary benefit of the tools
• Critical Thinking — the educational goal students feared losing
• AI Literacy — awareness and digital literacy as the paper's first recommendation

Connected Articles

• Cheating or not cheating? Rethinking AI-giarism and academic integrity through secondary students' ethical reasoning — "AI-giarism" and students' ethical reasoning about machine-assisted work
• Academic Integrity in the Age of AI: University Students' Study Practices and Ethical Judgments — university students' study practices and ethical judgments about AI
• Is using artificial intelligence tools for academic work cheating? Student perceptions, ethics, and the impact — whether students themselves call AI-assisted work cheating
• It's OK Because...": The Wild West of Student Rationalization of AI Use in Academic Writing — the reasoning students use to license AI-assisted coursework
• The psychological mechanisms and behavioral determinants of academic integrity in the age of artificial intelligence — behavioural determinants behind integrity judgments
• Perceptions Of Generative AI in the Global South: A Scoping Review — GenAI perceptions across Global South contexts
• Navigating the challenges of Gen-AI in Chinese higher education: Balancing technological innovation with academic integrity and intellectual engagement — a comparable policy-versus-practice tension in another system
• The Integrity of Psychology Assessments in the AI Age: A Critical Examination — what the assessment vulnerability looks like from the marker's side

Citation

Mulisa, F., & Mezgebu, T. (2026). Tools facilitate cheating, or partner supports learning? GenAI and academic integrity issues from students' perspectives. International Journal for Educational Integrity, 22(22).