How Universities Are Rethinking Student Assessment in the Age of Generative AI

The growing ability of large language models (LLMs) to solve complex problems and generate text, code, and other content has led to widespread use of generative artificial intelligence among students worldwide.

In a 2026 survey of 1,054 undergraduate students in the United Kingdom conducted by the Higher Education Policy Research Institute in Oxford, approximately 94% said they used generative AI tools to support assessment tasks. Around 12% admitted to inserting AI-generated text directly into their coursework.¹

In another study published in May, researchers analyzed survey responses from more than 95,000 students at 20 universities in the United States. They estimated that 9% of students had used AI in class despite knowing that doing so violated institutional rules.²

Educators around the world are increasingly noticing possible signs of AI-assisted work, including fabricated references, inaccurate citations, unusual punctuation, and writing styles associated with AI tools. Take-home assignments and online examinations have also raised concerns that generative AI may give some students an unfair advantage over classmates who do not use it.

“At one point, some students were submitting essays in which 50% of the references did not actually exist,” said Nikita Bezrukov, a professor of linguistics and communication at the Massachusetts Institute of Technology in Cambridge.

At the same time, many educators recognize the benefits of AI and believe students must learn how to use the technology responsibly. As a result, universities are rethinking traditional assessment methods to incorporate AI as an educational tool while addressing academic misconduct.

Educators around the world are experimenting with AI-focused assignments, supervised exams, oral assessments, process-based grading, and other approaches designed to evaluate genuine understanding.

“I don’t think anyone has the ‘right’ answer right now,” said Nicholas Mattei, a computer scientist at Tulane University in New Orleans, Louisiana. “Things are changing pretty quickly.”

Using AI to Develop Critical Thinking Skills

“For an assignment in my history of technology class, students identify information sources, summarize them using three LLMs, and critique the responses produced by each model. They then create an infographic and write an essay in which the use of AI is permitted.

Students earn additional points if they identify inaccuracies in the infographic or detect hallucinated references. The goal is to encourage students to think critically about what AI can and cannot do.”

— Nicholas Mattei, computer scientist at Tulane University

“Instead of asking students to write an essay about Adam Smith’s 1776 book The Wealth of Nations, I ask them to create a virtual representation of his economic theories using an AI agent that acts as a market trader.

We then examine different scenarios, including what happens when people can no longer trust one another. Students analyze the records created by the AI agents and write about their findings.

This approach makes it easy to identify whether a student has engaged with the activity. If they ask AI to complete every step at once, they often invent details instead of referring to the actual transcript.”

— Daniel Silver, sociologist at the University of Toronto Scarborough, Canada

Allowing AI in Assessments While Testing Methodology

“The examination for my web development course is open-book and open-internet, and AI is part of the internet. However, I assess a skill that AI does not have: the ability to build high-quality, reliable, and sustainable web applications.

I ask open-ended questions that can have multiple answers, such as ‘Why does this website not work?’ or ‘Why is this website slow?’ When students enter these questions into an AI chatbot, the tool often produces an answer without demonstrating a proper troubleshooting method. The response is also not necessarily accurate.”

— Ruben Verborgh, computer scientist at Ghent University, Belgium

“I do not assume that students are avoiding AI, so I approach assessment as if students and AI are working together. However, the tasks I assign cannot be completed effectively with ChatGPT alone.

I provide a screenshot showing a specific type of analysis involving neural activity data and ask students to describe what is happening. I also provide the assignment prompt and demonstrate how it might be entered into a chatbot.

If students use the AI output without evaluating it, the answer often has little connection to the screenshot. That becomes clear in their work.”

— Etienne Roesch, statistician and cognitive scientist at the University of Reading, UK

Asking Students to Document Their AI Use

“We ask students, particularly PhD candidates, who use AI in their research or writing to share their original interaction logs with AI tools.

The purpose is not simply to monitor students. These records help us determine whether students were deeply involved in the intellectual process, what knowledge was developed with the help of AI, and whether the technology supported student thinking rather than replacing it.”

— Yanjun Shen, researcher in ecology and engineering geology at Chang’an University in Xi’an, China

“Instead of submitting coursework as a PDF, students should use software with revision histories, such as Google Docs. This allows instructors to review earlier versions and understand how AI was used during the writing process.

I do not object to a student whose first language is not English using AI to improve grammar or clarity. However, students should not use AI to generate large sections of text.”

— Nikita Bezrukov, professor of linguistics and communication at the Massachusetts Institute of Technology

Reducing Take-Home Assignments

“In one course I teach, I redesigned the assessment so students had less writing to complete at home. Previously, students created two infographics. After seeing a large amount of AI-generated material, I replaced that assignment with an in-class pitch about higher education policies and programs, followed by a full presentation.

I now include more impromptu speaking activities and in-class writing assignments than I did previously.”

— Christina Ruiz Mesa, communication instructor at California State University, Los Angeles

“After moving away from traditional examinations, we returned to in-class written exams this spring. Students also write essays in class three times during the semester.

They may bring notebooks and books, but they must write in examination booklets. Mobile phones and laptops are not permitted.”

— Jennifer Sessions, historian at the University of Virginia in Charlottesville

Using Pass-Fail Grading to Encourage Genuine Learning

“For take-home data analysis assignments, I award only a pass or fail grade. There is nothing wrong with using AI to generate code if it saves time for more meaningful analysis.

Students may pass the course, but they can still struggle in the final examination, where they are permitted to use only a basic calculator. By removing detailed grading from these assignments, we encourage students to focus on learning.

We have never before seen so many students achieve perfect scores on an exam.”

— Natalia Sidalova, computer scientist at Eindhoven University of Technology, the Netherlands

“Although my department permits AI-assisted take-home assignments, some colleagues are reducing the weight of coursework and assessing learned skills through offline quizzes.

Some quiz questions relate directly to the assignments, which gives students an advantage if they completed the work themselves. We also use leaderboards to compare student performance and motivate students to engage with their coursework.”

— Maussama, computer scientist at the Indian Institute of Technology Delhi

Introducing Oral Examinations

“I ask students to write substantial amounts of Python code to interpret chemical data. Seventy percent of the grade is based on how the code works. The remaining 30% comes from a short, one-to-one oral examination in which students explain how specific lines of code function.

I do not mind if ChatGPT teaches a student a particular line of code. However, the student must understand and be able to explain every line they submit.”

— Gianmarc Grazioli, computational chemist at San José State University, California

“The written portion of the initial coursework I assign contributes only a small percentage of the final grade. The most important part is discussing the work and developing it further during a direct conversation with me.

During these discussions, students must demonstrate authorship and understanding. If they cannot explain their work or show proficiency, their grade is limited to a C.”

— Thea Goldring, instructor in the Princeton Writing Program at Princeton University, New Jersey

Using AI Detection Tools Carefully

“For assignments submitted in online classes, I use an AI detection tool provided by the university. It can help confirm some of my initial impressions, although it is not definitive.

Other warning signs include fabricated references and significant differences between a student’s first draft and final essay. I try to make a fair assessment based on several pieces of evidence rather than relying on a single detection tool.”

— Jacqueline Evans, psychology researcher at Florida International University in Miami

The Future of AI and University Assessment

Generative AI is changing how students learn, write, code, research, and complete academic assignments. While universities continue to address concerns about plagiarism and academic integrity, educators are also exploring ways to make AI part of responsible learning.

The emerging assessment strategies focus less on the final product alone and more on the student’s reasoning, process, documentation, communication, and ability to explain their work. In many courses, this means combining AI-assisted assignments with in-class writing, oral examinations, revision histories, practical demonstrations, and reflective analysis.

As AI tools continue to evolve, higher education institutions will need flexible assessment policies that promote ethical AI use while ensuring that students develop the knowledge and skills their courses are designed to teach.