The growing use of artificial intelligence in education has created a new challenge for universities and teachers as they try to distinguish genuine student work from assignments produced with tools such as ChatGPT. Johns Hopkins University professor Steve Hanke has offered his view on the issue, saying that experienced teachers can often notice when a student's work does not match their usual capabilities.
Hanke, a veteran professor of applied economics who has taught at Johns Hopkins for nearly 60 years, told Business Insider that he is not particularly worried about students attempting to use AI to cheat because he believes such attempts can often be recognised. He described himself as experienced enough to identify significant differences between authentic student work and material generated with the help of an AI chatbot.
According to Hanke, one of the biggest warning signs is a sudden improvement in the quality or sophistication of a student's writing. A student who normally submits basic or inconsistent work may suddenly produce an unusually polished assignment containing arguments, terminology or analysis that appear far beyond the level demonstrated in class.
Teachers who know their students well can also notice differences in reasoning. A written assignment may appear impressive on the surface but reveal weaknesses when the student is asked to explain the argument, defend a conclusion or answer follow-up questions. Hanke said his familiarity with students' abilities makes such inconsistencies easier to recognise.
This issue has become increasingly important as generative AI systems have become more capable. ChatGPT and similar tools can produce essays, summaries, explanations and other academic material within seconds. Their growing accessibility means students can use AI for legitimate learning support, but they can also use it to complete work that is expected to represent their own effort.
However, the question of whether AI generated writing can reliably be detected is more complicated than simply identifying a change in writing style. A student may improve naturally, receive outside tutoring, use grammar software or receive help from another person. Conversely, AI systems can now be instructed to imitate different writing levels and styles.
For that reason, a sudden change in writing should generally be treated as a reason for further discussion rather than automatic proof of cheating. Educators may compare the assignment with earlier work, ask students to explain their reasoning or require an in person assessment to establish whether the submitted material reflects the student's understanding.
The concerns surrounding AI assisted cheating also extend beyond academic rules. Educators worry that relying on AI to complete assignments can reduce opportunities for students to practise writing, problem solving and critical thinking. A recent analysis of student learning behaviour found evidence that some students can complete work more quickly with AI while learning less from the process.
At the same time, not all student use of ChatGPT is academic misconduct. Research and discussions among educators increasingly distinguish between using AI as a learning assistant and using it to produce work that students submit as their own. The American Psychological Association has noted that students generally recognise the difference between legitimate uses such as gathering background information and inappropriate uses such as having AI answer exam questions or complete assignments.
This distinction is becoming important as universities develop new policies for AI use. Some institutions allow students to use AI for brainstorming, explanation and editing, while requiring disclosure when substantial AI assistance is used. Other courses restrict AI use for particular assignments or require students to demonstrate their work through drafts, classroom discussions or oral assessments.
The debate also highlights the limitations of automated AI detectors. Educators have increasingly warned against treating detector scores as definitive evidence because AI detection tools can produce false positives and may perform differently across writing styles and languages. A teacher's understanding of a student's learning history can therefore be more useful than relying on a single automated score.
Hanke's argument is based largely on his experience as a long serving professor and his familiarity with student performance. His comments should therefore be understood as his personal assessment rather than a universal method for identifying AI generated work. Not every change in writing quality indicates cheating, and students who use AI tools may not always display obvious inconsistencies.
For students, the issue is also one of academic responsibility. Using AI to understand a difficult concept, generate practice questions or improve study habits may be permitted under a course's rules. Submitting AI generated answers as original work when such use is prohibited can result in academic consequences.
Universities are therefore moving toward a more nuanced approach in which the goal is not simply to detect AI but to design assessments that measure genuine understanding. Personalised assignments, classroom presentations, handwritten or supervised work and discussions about the submitted material are among the methods educators are considering.
The rapid development of AI means the relationship between students and academic institutions is likely to continue changing. Teachers may become more familiar with AI generated writing, while students may become better at editing or adapting AI output. This creates an ongoing cycle in which assessment methods will need to evolve.
Hanke's comments have nevertheless added to a broader discussion about whether experienced educators can recognise AI assisted work from inconsistencies in writing, reasoning and subject knowledge. His view is that teachers who understand their students' normal capabilities can notice when an assignment does not appear authentic.
Ultimately, detecting AI use is not simply a matter of spotting unusual vocabulary or polished grammar. The more reliable approach is to assess whether the student can demonstrate the knowledge and reasoning contained in the work. As AI becomes more common in education, universities are likely to place increasing emphasis on that underlying understanding.