The new plagiarism checker recognises Ghost writing and ChatGPT

The new plagiarism checker recognises Ghost writing and ChatGPT

Should we ask students to take an oath?

19-09-2023 · Interview

Bad news for students who have put their bets on ChatGPT. From September onwards, UM will be using a new plagiarism detector, which can also detect AI. But how great are the chances of being caught?

Since 2020, UM has worked with the plagiarism checker Ouriginal, but many lecturers and Examination Committees complained about it. It was never proven beyond doubt, but it was suspected that Ouriginal did not meet requirements, says Charles Bollen, co-ordinator of UM’s digital learning environment. “This plagiarism checker missed a couple of cases of plagiarism, because it did not have access to the journals from which the texts were copied.”

Another disadvantage of Ouriginal is its lack of user-friendliness. “Checking a paper required a lot of clicking,” says Maarten van Wesel, plagiarism expert at the university library. “That takes a gigantic amount of time when you have hundreds of papers to check.”

With the UM’s new plagiarism checker, Turnitin Originality, Van Wesel believes that things will be a lot faster. "This software is from the American company of the same name that is dominating the market at the moment. After a tender procedure, UM signed an agreement with this company. From September onwards, Turnitin Originality will be used to check all papers and theses for plagiarism."

Trade secrets

Plagiarism is defined as copying one’s own or someone else’s ideas or words without any reference, so without mentioning the source. Last year, more than 200 UM students were guilty of this. Using ChatGPT or other forms of artificial intelligence, it is strictly speaking not a case of plagiarism, because an original text is being generated. But this is also impermissible because it was not written by the student.

Turnitin Originality is capable of detecting texts from ChatGPT and other AI. “According to the company, the software is very good at doing so,” says Van Wesel, “but this should become apparent when we use it in actual practice. Chances of it succeeding, are at any rate greater the more AI has been incorporated in the text.”

How the detector goes about detecting AI, is a trade secret, says Bollen. “The producer is not revealing much.” It is known, at any rate, that an analysis of the sentence structure plays a role.

The question is, however, how important is the detection of AI, if the majority of lecturers do not want to prohibit chatbots, but rather integrate them into the curriculum. “That is correct,” says Bollen. “As far as I know. there is not a single UM faculty that wants to ban AI. If you use ChatGPT as an aid and write your own story afterwards, you can learn from it. Although AI texts are not always correct.”


Students will also no longer get away with ghost writing, which is handing in a paper that was made by someone else. Or from a so-called paper mill, where you can order and pay for a paper or thesis.

A former student of Arts and Social Sciences used different alphabets in her thesis. She used Cyrillic letters such as the P and the R, which are also in the western alphabet. Ten years ago, you could fob off the plagiarism software by doing this, but that is also a thing of the past. Turnitin Originality can detect this.

Researchers have their own plagiarism checker, called iThenticate, also from the Turnitin company, which allows them to check their articles (and theses) themselves before they submit them to journals. Students are not permitted to do this. Why not?

Bollen: “They would then be able to check beforehand whether they are going to get caught. If it doesn’t detect anything, they can safely plagiarise.” 

Van Wesel: “The check is more understandable for researchers. They often work in teams with colleagues living abroad, whom they don’t know very well. By checking, you can prevent problems caused by someone else’s mistake. Researchers who write a lot, can check whether they are not guilty of self-plagiarism. That can easily happen when you continually publish about the same subjects.”

Literature references

There are also disadvantages to Turnitin Originality. This detector is twice as expensive as Ouriginal, and much more expensive than SafeAssign, which UM used for a long time in the past. SafeAssign was free of charge, because it was part of Blackboard, but had another huge advantage. If the plagiarism score was 10 per cent, this meant that chances of plagiarism in this paper was 10 percent. Anyone who had to check four hundred papers, could accept this paper with an easy mind.  

Turnitin Originality (and Ouriginal) works differently. If it shows 10 per cent, this means that 10 per cent of the text is plagiarised. Van Wesel: “You cannot just leave such a paper unchecked, because maybe the entire conclusion has been copied literally.”

Bollen: “During the negotiations for tender, we had additional discussions with Turnitin about this. We wanted to know if the percentages could be presented in the style of SafeAssign, but that wasn’t possible.”  


Are the chances of being caught higher with Turnitin Originality? “I could not say,” they both say in chorus. Van Wesel does think that “it does have a greater deterrent effect”. “Compare it to traffic lights. When there is a police camera close by, you will think twice about crossing the line.”

Researchers have to swear an oath, in which they publicly promise to abide by the rules of academic integrity at all times, before they embark on a PhD, says Bollen. “Maybe we should ask the same of students. It already happens at some universities.”

Van Wesel: “But the question is whether it will help. There are two kinds of plagiarising students. Those who have no moral problem with committing fraud. This group won’t pay much heed to an oath. And there are students who cross the line due to lack of time or stress. The latter group has other problems and will probably feel extremely guilty."

Illustration: Bas van der Schot

Tags: plagiarism, turnitin, chatgpt, instagram

Add Response

Click here for our privacy statement.

Since January 2022, Observant only publishes comments of people whose name is known to the editors.