Thousands of UK university students have been caught misusing ChatGPT and other artificial intelligence tools in recent years, and while traditional forms of plagiarism have shown significant decline, the Guardian investigation reveals.
The academic integrity violation investigation amounted to 5.1 per 1,000 students in nearly 7,000 proven cases of fraud using AI tools between 2023 and 24. This is up from 1.6 cases per 1,000 people in 2022-23.
Experts say the biggest figures will increase this year and again to around 7.5 proven cases per 1,000 students, but recorded cases only represent the tip of the iceberg.
This data highlights the rapidly evolving challenges for universities. It is attempting to adapt evaluation methods to the emergence of technologies such as ChatGPT and other AI-powered writing tools.
Plagiarism accounted for almost two-thirds of all academic misconduct before the widespread availability of generative AI in 2019-20. During the pandemic, plagiarism intensified as many ratings moved online. However, as AI tools became more refined and accessible, the nature of fraud has changed.
The survey predicts that according to the initial digits for this academic year, confirmed cases of traditional plagiarism will fall from 19 per 15.2 to 15.2 in 2023-24, down to about 8.5 per 1,000.
The Guardian has contacted 155 universities under the Freedom of Information Act, which requires numbers for proven cases of academic misconduct, plagiarism and AI fraud over the past five years. Of these, 131 Although we provided some data, not all universities had records of annual or fraud categories.
Over 27% of responding universities did not record AI misuse from 2023-24 as another category of fraud. This suggests that the sector is still aware of this issue.
Many cases of AI fraud may not be detected. A survey by the Institute for Higher Education Policy in February found that 88% of students use AI for their assessments. Last year, researchers at the University of Reading were able to test their own rating system and submit the work generated to AI without detecting 94% of the time.
Dr. Peter Scarf, an associate professor of psychology at the University of Reading and co-author of the study, said there have always been ways to cheate, but the education department must adapt to AI, creating a fundamentally different problem.
He states: “I imagine the catch represents the tip of an iceberg. AI detection is very different from plagiarism that allows you to see copied text. As a result, in situations where you suspect AI use, it is almost impossible to prove that you don't want to use a student, if the AI Detector says (if you use it).
“It is impossible to simply move all the ratings students take in person. At the same time, the sector must admit that students are using AI even if they are not asked and detected.”
Students who wish to be undetected using generative AI have drawn a lot of online material. The Guardian found dozens of videos for students about Tiktok Advertising AI paraphrases and essay writing tools. These tools help bypass common university AI detectors with the “Humanising” text generated by CHATGPT.
“It's extremely difficult to prove the misuse of AI by students who are well used to it and know how to edit the output,” said Dr. Thomas Lancaster, an academic intensity researcher at Imperial College London.
Harvey* has just finished his final year of earning a degree in Business Management from Northern University in English. He told the Guardian that he used AI to generate ideas and structures for the task and proposed references, and that most people he knows used the tools to some extent.
“When I first joined Uni, ChatGpt came in, so it was always there for me,” he said. “I don't think a lot of people use AI and then copy it to fit the word. I think it's just to help brainstorm and create ideas in general. Anything I take from it, I'd completely remake it in my own way.
“I know who used it and then use other methods of AI to change it and write AI content so that it sounds like it's coming from humans.”
Amelia* just finished her first year of her music business degree at a university in the southwest. She also used AI to use summary and brainstorming, but said the tool was most useful for those with learning difficulties. “One of my friends use it to structure her essays rather than writing for her or studying anything, instead putting their own points in and doing some. She has dyslexia.
Science and Technology Secretary Peter Kyle recently told the Guardian that AI should be deployed to “level up” opportunities for dyslexia children.
It appears that technology companies are targeting students as a key demographic for their AI tools. Google offers college students a free upgrade of their Gemini tools for 15 months, while Openai offers discounts to college students in the US and Canada.
Lancaster said: “University-level assessments can seem meaningless to students, even if there are good reasons to set this up as educators. All this will help students understand the reasons needed to complete a particular task and become actively involved in the assessment design process.
“There are often suggestions that more exams should be used instead of written assessments, but the value of memorization learning and retained knowledge continues to decline each year. I think it's important to focus on communication skills, people's skills, skills that cannot be easily replaced by skills that cannot be easily replaced by skills that cannot be engendered in new skills and confidence to succeed in the workplace.”
A government spokesperson said it has invested more than £187 million in the national skills programme and has released guidance on the use of AI in schools.
They stated: “Generating AI has great potential to transform education and offers exciting opportunities for growth through planning for change. However, integrating AI into education, learning and assessment requires careful consideration, and universities need to decide how to leverage their benefits and reduce risks to prepare them for future work.”
*Name has been changed.
