The biggest risks for misuse of AI in assessment lie in unsupervised assignments, especially written assignments. It is, of course, not possible to completely avoid unsupervised written assignments, nor would we always want to do so. We need to reconsider whether they still fit the learning objectives at a time when many students are using genAI, such as ChatGPT, to help them.

Important: Do not rely on plagiarism detection software

The existing methods of detecting AI-generated texts are unreliable and should not be used. It is unclear if this situation will change, as AI is constantly learning and improving, leading to future versions with improved functionalities, making unauthorized use even more difficult to detect. In addition, there are privacy concerns about detection software that can recognize students’ writing styles.

Instead of using detection software, we recommend that lecturers take a closer look at their written assignments. It might be that alternative methods of assessment can be used or that the assessment for the course as a whole needs to be reconsidered. If so, have a look at our page on assessment at a course level.

As previously mentioned, it might not be possible to avoid using unsupervised written assignments, such as papers, take-home exams, or coding assignments, in your course at the present time. In that case, take a look at the steps below that help you evaluate the position of your written assignment. You will also find practical tips and examples on making your written assignment less vulnerable to the misuse of genAI.

Note that it is impossible to set up unsupervised assignments in such a way that misuse of genAI is prevented entirely. Students can be very adept at integrating genAI at different stages of the writing process in ways we can neither anticipate nor detect. It is better to focus on how you can motivate and stimulate students to engage in the learning process responsibly.

Step 1: consider alternative assessment forms

Take a closer look at the learning objectives of the course and consider whether assessment forms other than unsupervised written assignments can be used to determine whether students have achieved those learning objectives. Another consideration is whether you can reduce the weight of the unsupervised writing assignment(s) in the final grade in favour of other assessments less prone to the unauthorized use of AI.

Also consider the place of the course within the curriculum and consult with the rest of the teaching team. Many programmes include a writing assignment in every course, even though this is not always necessary or desirable from the perspective of the learning objectives and learning outcomes. In which courses are writing assignments really indispensable? Are there courses in which a different form of assessment can be used? See for more information our page on assessment at a course level.

Step 2: consider the assignment requirements

Take a close look at the assignment requirements. We have identified four particular vulnerabilities in assignments that can often be resolved. Read more about these vulnerabilities, solutions, and detailed examples by clicking the links below.

Additional ways to make the assignment guidelines less susceptible to the unauthorised use of AI

Submit the assignment description to ChatGPT or another genAI tool, possibly with some additional prompts, to see whether their responses are acceptable, similar to student work, and could perhaps even achieve a passing grade.
Ask for reflection on the argumentation of author(s) of books, articles, etc. Have students compare different sources and evaluate arguments.
Ask for creative solutions, interpretations, and connections, and/or the use of diagrams, graphs, images, etc.
Ask students to reflect on their personal experiences and opinions (for instance: “Explain how last week’s materials have changed what you think about the role of…”)
Supervise students during parts of the written assignment. Ask students to do part of the writing in class, for instance, by organizing brainstorming sessions or peer feedback sessions. You can also organize hands-on demonstrations based on students’ written work, such as interpretations of primary sources or artefacts, mock therapy sessions, laboratory settings, or code demonstrations.
Focus on the process rather than the final product. Build steps into the assignment where students reflect on their progress (self-evaluation), either orally or in writing, and give feedback on this or include it in the assessment. Shorter, more frequent (formative) assessments during class may also be an option. Specific options are:
- Have students keep a diary or logbook to show their approach (and consider including this in the final grade).
- Have students explain their thought process (with track changes, for instance).
- Give students the option to document their writing process in different ways (for instance, drawings, mind maps, audio recordings).
- Organize progress meetings/presentations throughout the process and assess draft assignments.
- Have students make a plan for their assignment together during a group session.
- Interview students about their products.
- Gain insight into students’ notes during the research/writing process.

Step 3: adapt your teaching to the new situation

Try to keep track of the writing process during the teaching block (see also previous step), e.g. by having students hand in draft products, discussing the approach with them, having them carry out part of the writing process on location (within working groups), and/or give a (verbal) explanation of their (writing) process. Give feedback as much as possible during the process (rather than on the final product). Consider using peer feedback in small groups.
Appeal to students’ motivation for their studies. Explain that making an effort is important for learning and that shortcuts with AI can limit your longer-term intellectual growth. A lot of students turn to AI because they are doing things they don’t enjoy doing or see the point of. Other factors that can play a role are uncertainty, stress, pressure to get high grades, and a lack of time. Addressing these concerns will motivate students to learn and empower them to trust their knowledge and skills, so they’ll be less likely to rely on AI.
Discuss ethical and practical objections to the use of AI with your students:
- Discuss what academic integrity means in your own field.
- Give examples of academic dishonesty.
- Explain that not only fraud but also unreliable AI output is unacceptable in academic work.
- Make your standards explicit: discuss what is allowed and what is not. Distinguish AI-use from AI-misuse.
- Show what you gain from doing your own research.
- Encourage students to submit questions or dilemmas about integrity.
- Use peer reviews to motivate students to hold each other accountable for academic integrity.
- Ask students to sign an authenticity statement upon submission.
- Show how you can use AI in a meaningful and ethically acceptable way. Have students keep a log of their usage.

Step 4: zoom in on the assessment itself

If you have doubts about the unauthorized use of AI impacting your assessment, consider the following points:

Change the weighting in the rubric to emphasize aspects of writing relating to higher level skills that cannot be done well by genAI, and decrease the weighting of aspects that AI can do well (such as spelling, structuring, summarizing, defining, etc.). However, in doing so you must be careful not to bias your assessment in favor of students who use genAI versus those who do not.
In some courses you might need to assess precisely these foundational academic writing skills. In this case, changing the weighting is not an option, and it might be better to have students write under supervision.
Make sure your assessment does not only include the final written product. E.g. include the tips mentioned above in the assessment, such as oral presentations of the written work and the process that led to it, self-evaluation, peer feedback, etc.
It is impossible to determine reliably whether a text was generated by AI. But there are some indications you can look out for:
- A flat, generic style – AI is good at mimicry, not originality.
- In languages other than English: unusual language, strange mistakes, etc.
- Statements that strongly diverge from the course material.
- Incorrect, overly generic, or missing referencing.
- Unexplained discrepancies in text length.
- Inconsistent tone or style.
- Lack of personal opinion or reflection.
- Similarities between your own AI-generated text and the student’s text.
- Inability of a student to answer (simple) questions about their own text.
Do you suspect that a student used genAI to commit fraud or plagiarism in your assignment? Then report this to the Examination Board so that they can investigate further and sanction it if necessary. You can discuss your suspicions with the student, but suspicions alone do not serve as evidence in fraud cases.
Note that the development of genAI is rapid. What seems to be a weakness of AI now, might be improved tomorrow. So don’t rely too much on these weaknesses or on your own ability to detect students’ use of genAI.

Advice on making your assessment safer

Would you like advice on making your assessment safer? Please contact an assessment specialist through your faculty’s TLC. You can also always ask the TLC-Central’s assessment specialists for advice by contacting tlc@uva.nl.