University Flags 95% of Theses for Plagiarism

The University of Lucknow is a school based in Lucknow, India. According to Wikipedia, the school has over 20,000 students, including over 6,000 postgraduate students.

The Times of India reports that the school’s library, Tagore Library, recently performed an analysis of some 121 theses that had been submitted to the school. Of the ones checked, 116 were determined to have significant evidence of plagiarism, representing over 95% of the submissions.

Under the school’s rules, once a student finishes their thesis, they submit a soft copy to the library. That copy is then checked using DrillBit, the school’s chosen anti-plagiarism program, which looks for both traditional copy and paste plagiarism as well as AI usage.

According to those policies, when plagiarism is detected in a thesis, it is returned to the student for corrections.

Professors at the school applauded the detection, saying that it rewards students who put in the work and serves as a counterbalance to the overreliance on AI bots. However, under the surface, there are a lot of difficult questions.

That’s because a 95% detection rate is incredibly high and, most likely, points to a serious problem beyond just the issue of AI or plagiarism. It’s likely that something much more fundamental is amiss.

A Nearly Impossible Number

Though it’s virtually a guarantee that the rate of plagiarism/AI usage far exceeds the amount that is detected, a 95% detection rate, especially among PhD theses, is pretty much unheard of.

The first place that most people will look, including myself, is at the software used to make the detection, DrillBit.

Back in January, a technical issue prompted scholars at the school to express concern over the tool, saying that it inaccurately flagged plagiarism in legitimate work. The school said it had not received any formal complaints.

DrillBit is a relatively new application and is not one that I have used. As such, I cannot speak on it directly. That said, the school has said that it will be adopting Turnitin across all levels in the future, but this particular examination was done using DrillBit.

Given the extremely high detection rate, it seems likely that at least some of it is due to DrillBit falsely labeling original work as plagiarism. It’s unclear what percentage of the papers were flagged for traditional plagiarism vs. unauthorized AI usage. But given that DrillBit offers both, there are plenty of opportunities for false positives.

But even if half of these are false positives, that would still leave nearly half of the theses containing worrying amounts of plagiarism. Is it possible that something else is going on?

A Problem of Policy

On the University’s website, their anti-plagiarism page makes it clear that there is no percentage threshold for what is considered acceptable.

While it is a reasonable policy, this page is also clearly out of date. It references Urkund, a plagiarism detection tool that was renamed in early 2023 and shuttered mid-2025. It has not been available for nearly a year and has not been known as Urkund for over three.

Instead, the coverage of Lucknow has highlighted several different policies that use hard word count and/or percentage limits. For example, the story about this mass flagging claims that “According to Lucknow University norms, 180 words of plagiarism or less than 5% is allowed for citations and facts which are available in journals, books and even in AI databases.”

Meanwhile, the December 2025 story about the university acquiring Turnitin says that, “Under the newly implemented policy, students will no longer be able to submit assignments if the plagiarism level exceeds the permissible limit of 8-10%.”

Both of these quotes set off alarm bells in my mind.

Simply put, there is no percentage that indicates what is and is not acceptable when it comes to plagiarism. Plagiarism is about attribution and citation, not what is and is not copied. It’s possible for a paper with 0% copied content to be a work of plagiarism, and it’s equally possible for a work with a high percentage copied to be completely cited.

The percentage totals may give an indication as to where further investigation is needed or call for better paraphrasing practices, but they don’t indicate plagiarism by themselves.

A bad policy that sets hard limits risks both ignoring significant cases of plagiarism and turning legitimate scholarship into a plagiarism matter. There’s no mathematical formula for determining plagiarism.

I have to wonder how many students may have run afoul of whatever policy they were held to, even though their scholarship was original and ethical.

Bottom Line

To be clear, I don’t know what happened, and I can’t know what happened. The news coverage of the story is relatively thin and, even if it weren’t, I wouldn’t be able to offer much commentary without doing an evaluation myself.

But one thing is very clear to me. If 95% of the theses you examine come back as plagiarized, this is not a “send the papers back to the students for correction” moment. This is a “something has gone seriously wrong” moment.

The best case scenario is that it’s simply the software over-reporting or the policy flagging original work as plagiarism. That’s because the alternative is that 95% of your PhD candidates are committing plagiarism. That is a far more terrifying possibility.

While I fully agree that the amount of actual plagiarism, in particular AI-based plagiarism, is under-reported, this is an outcome that should be setting off alarms, not mild annoyance.

If the policy is bad or plagiarism really is that rampant, moving to Turnitin isn’t going to help. Though better tools can generate better results, that’s only if the foundation they are working from is solid. The best software can’t fix bad policy or bad instruction.

Regardless of why so many papers were flagged, the school needs to take this issue seriously and figure out what went wrong. Whether it’s the software, policy or the students, it needs to know the issue before it can hope to fix the problem.