There’s a lot of concern that OpenAI’s ChatGPT could help students cheat on tests, but how well would the chatbot do if you asked it to write a higher education exam? He’d pass it – albeit narrowly, Engadget wrote on the subject.

In a recently published study, law professors at the University of Minnesota had ChatGPT prepare the answers for graduate exams in four courses at their school. The AI passed all four, but with an average grade of C+. In another recent article, Wharton School of Business professor Christian Terwish found that ChatGPT passed a business management exam with a grade of B to B-. You wouldn’t want to use technology to impress academics, then.

The research teams found that artificial intelligence is inaccurate to say the least. A group from the University of Minnesota noted that ChatGPT is good at dealing with “basic legal rules and laws” and summarizing doctrines, but struggles when trying to point out issues related to a case. Tervish said the generator is “amazing” at simple operations management and process analysis questions, but cannot handle advanced process and reasoning questions unique to humans. He has even made mistakes with 6th grade level math.

There is room for improvement. The Minnesota professors said they did not tailor the text-generation prompts to specific courses or questions and thought students could get better results with customization. ChatGPT couldn’t handle an exam or essay on its own, but a cheater could make the system generate rough answers and refine them.

Study groups still think ChatGPT could have a place in the classroom. Educators could train students how to rely on AI in the workplace or even use it to write and grade exams.