Launched in 2022, ChatGPT quickly took the world by storm. By August 2025, it was already being used by 700 million people every week – almost as many people as WhatsApp has monthly users!
From writing emails and helping with difficult decisions to answering intriguing questions, ChatGPT is now seen as a useful tool for nearly every aspect of our lives. Students, too, have embraced the potential of generative artificial intelligence (gAI).
In February 2025, the British Higher Education Policy Institute (HEPI), in collaboration with Kortext, published a report based on a survey of 1,041 undergraduate students. The findings are striking: 92% of respondents said they use gAI tools, up from 66% in 2024, when the survey included 1,250 students. What’s more, 88% said they had used such tools to prepare assignments graded by instructors – up from 53% the year before. At the same time, the share of students who said they did not use gAI at all dropped from 47% in 2024 to just 12% in 2025.
There are plenty of reasons why tools such as ChatGPT have caught on among students. Saving time and effort – and, in students’ view, producing higher-quality assignments – are just some of them.
But the use of artificial intelligence in education also raises a host of concerns, particularly ethical ones. It can lead to academic misconduct – for example, when a student submits a paper prepared entirely or largely by gAI as their own work. As this practice becomes increasingly common, it presents a serious challenge for academic instructors.
Dr. Krzysztof Skonieczny of the Faculty of “Artes Liberales” at the University of Warsaw and Dr. Valerija Vendramin of the Educational Research Institute in Ljubljana examined the issue by asking Polish and Slovenian lecturers how they view students’ use of ChatGPT – and how they believe unethical practices can be countered.
What constitutes cheating?
The survey, conducted via an online questionnaire in January and February 2024, included 272 faculty members: 152 from Polish universities and 120 from Slovenian universities.
When asked whether students’ use of ChatGPT to prepare written assignments is dishonest, 49 respondents answered, “Yes, always,” while 19 said “No, never.” The vast majority believe that using such tools is dishonest “in some cases.”
The next question asked respondents to identify which ChatGPT practices they consider dishonest. A total of 185 people responded. Based on their answers, the researchers identified three types of dishonest use: (1) ChatGPT does all or a significant part of the student’s work – 80 respondents; (2) Students present ChatGPT-generated work as their own or fail to disclose that they used the tool – 50 respondents; (3) Students use ChatGPT to complete specific tasks, such as writing articles or essays, translating text, or taking exams – particularly when completing theses or term papers – 41 respondents.
Is it possible to detect cheating?
The second area of the study focused on how academic instructors try to counter students’ dishonest practices. In a closed-ended question, respondents rated the effectiveness of different methods on a five-point scale, from least to most effective, and indicated whether they used them themselves. An open-ended question also allowed them to suggest solutions not included in the survey.
According to respondents, the most effective strategy is discussing the student’s work with them. As many as 175 respondents reported using this approach: 100 from Polish universities and 75 from Slovenian universities.
Another effective method is monitoring students’ work at different stages of development. 98 respondents reported using it. Here, however, there was a notable difference between the Polish and Slovenian groups: 69 Polish faculty members versus 29 Slovenian faculty members. The researchers note that the study cannot explain the reason for this difference.
Explaining which uses of ChatGPT are dishonest was also considered effective. This approach was used by 133 faculty members: 71 from Poland and 62 from Slovenia.
Other methods included: changing written assessments to oral ones – 73 teachers (38 from Poland and 35 from Slovenia); assigning work to be completed in class rather than at home – 65 respondents (39 from Poland and 26 from Slovenia).
By contrast, specialized AI-detection software is rarely used. Only 26 respondents – 14 from Poland and 12 from Slovenia – reported using such tools to check assignments. Research shows, however, that their effectiveness is limited: they can incorrectly flag texts written entirely by humans as AI-generated.
Finally, punishing students for using ChatGPT was considered ineffective. Only 20 respondents – 10 from each national group – identified punishment as a measure they use to combat dishonest practices.
Using AI without crossing the line
A study by Dr. Skonieczny and Dr. Vendramin shows that academic instructors largely agree on what counts as unethical use of artificial intelligence.
Instructors see tools such as ChatGPT as support tools – not substitutes for students’ own work. They can help refine the style or language of a text, for example. But students must remain the actual authors. They must write the texts themselves, while the ideas, concepts, and arguments must come from their own intellectual effort.
“It seems to me that we should rather aim to act with honesty and as consciously as possible. If we decide to use ChatGPT or another tool to write a text, we should do so only to the extent that it does not undermine the purpose the text is meant to achieve.”
“Ted Chiang recently made a similar point in an article for The New Yorker, comparing writing student papers with ChatGPT to bringing a forklift to the weight room: we may be able to lift any barbell effortlessly, but we will not build muscle mass,” notes Dr. Krzysztof Skonieczny.
This points to the real issue: intellectual effort. Using tools such as GPT means taking a shortcut – and that shortcut can very easily make students feel they no longer need to think. But intellectual effort is at the heart of learning. It is what allows students to develop cognitive skills and the ability to think critically. And, as some studies suggest, the increasingly frequent use of AI tools has a negative impact on cognitive abilities.
The study’s authors also point to the crucial role of academic instructors in this new academic reality.
“Personally, I prefer to think of generative artificial intelligence as a challenge that – perhaps paradoxically – forces us to improve the quality and intensity of the relationship between academic instructors and students.”
“The easiest way to think about this is probably in relation to the topic we address in the article, namely potential cheating using gAI applications,” explains Dr. Skonieczny.
“We currently do not have – and likely will not have – a fool-proof, 100% effective way to detect the use of artificial intelligence tools to generate text. So if we want students to write their own texts anyway – or at least without breaking the rules we’ve agreed upon – we must be able to convince them to do so. That means convincing them that the assignments we give are meaningful and important.”
“To do this, we need students to trust us as teachers, the procedures we follow, and the educational institution or university itself. We can build such trust only through a thoughtful relationship; and only when we ourselves clearly understand why we assign these particular tasks rather than others,” the researcher emphasizes.
Perhaps this is where the greatest challenge lies: instead of chasing students and searching for perfect AI detectors, we need to make the meaning of studying stronger than the temptation to take shortcuts.
The text was originally published in Polish on the Serwis Naukowy UW website on September 22, 2025.
