ALP Guide: Assessing and developing student writing skills in the age of AI

Assessing and developing student writing skills in the age of AI

• Should my assessment be writing based? 

  The first question you should ask yourself is whether you want your assessment to be writing based or not. Here, the writing pie can help you think about what your learning goals actually are when it comes to writing, and whether a writing-based assessment is the most suitable approach.

  Writing is often used to assess student knowledge, and the left side of the writing pie depicts skills related to knowledge, content, thinking, and the process of writing.  Here, it’s important to question whether a take-home writing assessment is the best way to assess knowledge for your course, given the ease with which AI writing tools can create informational texts mimicking knowledge. In this situation, you could consider controlled assessments where it is not possible for students to access AI tools, for example in oral exams or an on-campus exam.

  Here are some specific questions that can help you evaluate whether you want your assessment to be writing based.

  • What are the learning goals of your course? Do they include learning to write?
  • Is writing the best format for students to demonstrate their learning?
  • Is your course designed to help students improve their written communication?
  • Do you provide feedback on writing skills?
  • How vulnerable is your assessment to AI tool misuse? See the section below for a guide to how you can check this.
  • Could you redesign the assessment to make it less vulnerable? 
  • Could you redesign the assessment so that it only focuses on the learning goals that you want to evaluate? For example, if you focus on understanding and paraphrasing, you could allow students to write in a more personal and informal way.  
  • Would you like your students to learn to write with the aid of AI, as well as without it?

  How vulnerable is your assessment to AI tool misuse? 

  One intuitive and simple suggestion to help you identify the kind of responses an AI tool might generate in response to your assignment is to actually try to create responses yourself. Follow the steps below to do this:

  1. Take your assignment guidelines and rubric and any other student-facing material into the AI tool and ask it to generate a response. Try using the kind of prompt your students might use
  2. Do this several times and use different AI tools like ChatGPT, Claude or Copilot to see how their responses differ.
  3. Look at the result and ask yourself some of the following questions:
     • How does this differ from student texts I would expect to see?
     • Does this resemble any of the student texts that I have had doubts about?
     • Where does this match or not match the instructions, rubric, and in-class materials?
     • What kind of language and topics does the AI tool come up with? Are they relevant to the course? Or are they overly general statements? Are they the kind of statement that you might expect from a student, or are they more reminiscent of a synthesis article from an experienced academic?
     • Does the paper make overly strongly claims and statements without much foundation?
     • How effectively does the paper deal with terminology in the field? Is there any inconsistency with the use of terminology, or is it poorly explained?
     • Is the analysis strange?
  4. Students may use various follow-up prompts to improve the text. Play around with different prompts yourself to see how it changes the result.

  Once you see the outputs, you may want to consider revising your assignment prompt to make it less vulnerable to AI.

  Alternative assessments

  If you decide that you would like to use an alternative form of assessment, you might consider some alternative writing formats such as:

  • real-world problem-solving tasks and scenarios such as case studies or simulations based on industry practices (Evangelista, 2025);
  • multimodal projects such as podcasts, video essays, poster presentations;
  • portfolios.

  You might also consider alternative formats that include no writing:

  • oral exams such as presentations and live Q&As;
  • performance-based evaluations such as evaluating skills in the classroom or practical exams (Evangelista, 2025);
  • project-based learning.

  It is still important to note that many of these alternative forms of assessment are not completely “AI proof”. For example, Liu (2024) demonstrated how AI tools could be used to effectively complete multimodal assessments.

  If writing is a learning goal

  If training students to write is in fact one of your learning goals, the writing pie can help you think about which aspects of writing your learning goals relate to. This can help you redesign your assessment, by allowing you to focus on the learning goals that you really care about. For example, you might want students to learn how to effectively structure an academic text. Then, you could have students analyse an article in their field and write about what they learn. You could allow them to write more informally and reflectively so that they can focus more on the structure and worry less about following academic conventions themselves.

• How can I make my assessment less vulnerable to AI tool misuse?

  Although it is not possible to create a completely “AI proof” assessment, there are several possible approaches to make your assignments more engaging and motivating for students and hence reduce the likelihood that they will over-rely upon AI.

  Tips to make your assessment less AI vulnerable

  • Break longer projects into smaller steps: Rather than focusing on assessing one large piece of work at the end of a project, divide larger projects into smaller sections. This makes the whole project more manageable and less intimidating for students.  These smaller steps could be formative assessments that are graded as pass/fail so that these steps are less about assessing the students, and more about supporting them in the writing process. This can scaffold student learning and also provide a history of the text, which could be also useful in detecting possible AI tool misuse. You could focus on individual elements of the writing pie. For example, a student could use a text plan to show that they can divide a text logically. You can also include writing conferences, and opportunities for peer and teacher feedback in these smaller steps (Savini, 2025). Feedback can give students a lot more confidence in themselves, but it doesn’t have to take a lot of work from you. The feedback could be on short sections of the text, for example 1 or 2 paragraphs early in the assignment to check that the students are on track.
  • Use in-class hand-written assessments: You could include a short in-class assessment that involves the students writing some work by hand with no (or limited) access to technology. This allows you to assess their writing independent of aid and gives a sample that you can compare with later. This might act as a deterrent because students are aware that you have a sample of writing they completed without access to AI writing tools. You could allow the students to complete some preparation work before the session. This could make the writing less threatening and allow students to focus more on the ideas and content.
  • Create authentic assignments: You can make writing more authentic and meaningful in a variety of ways: you can focus on authentic genres with real-world audiences, share writing beyond the classroom, address issues that are important to students, and relate the writing to their future goals (Savini, 2025). Meaningful and authentic assessments can motivate students to write independently by emphasizing how completing the assignment themselves benefits them. For example, instead of an essay, business students could write a Linkedin article on early-career leadership strategies; engineering students could draft a renewable energy design proposal for the university; and health sciences students could create a health awareness campaign on a topic relevant for students.
  • Check that assignment instructions are accessible and engaging: Savini (2025) notes that assignment instructions are sometimes written in a manner that can be intimidating for students, with a large amount of dense text and a focus on requirements. Reframing assignments as “invitations” (Savini, 2025) could make them more motivating for students. This could be achieved by emphasising the aims of the assignment and de-emphasising grammatical accuracy, academic style and formatting requirements. The Transparency in Learning and Teaching in Higher Education Framework provides some useful examples of accessible and engaging assignment instructions.
  • Include personalised/reflective elements in the assignment: You can ask students to include personal experiences and reflection in their writing assessment, encouraging students to engage more with the assignment. Incorporating personal experiences in writing should also mean that the texts include more unique and personalised details, whereas an AI tool is more likely to produce more generic responses and Goulart et al. (2024) found that AI tools produced more formal sounding writing when prompted to write assignments requiring personal information. However, Li et al. (2023) found that ChatGPT was capable of generating convincing reflective writing. So, it is important to think carefully about the prompts for personalised and reflective elements.
  • Relate the assessment closely to the classroom content: You can require students to include specific tools and frameworks that have been used in the classroom. If students don’t use these, then it can be an easy marker that the student hasn’t followed the assignment instructions or that the text is AI generated. This approach is especially useful if these tools and frameworks are specific to your course, or even if you have approached more common tools in a particular way.
  • Incorporate group work: Including group-work in assignments may encourage students to self-police the use of AI tools among their group. Some group members may be entirely against the use of AI tools, and some may worry about overuse of the tools. Group work can also make assignments less individually threatening for students because they can work with their peers and check their ideas together. This could mean that they are less likely to turn to AI tools because they are more confident in their own ideas.

• How can I help my students develop writing skills in the age of AI?

  The easy accessibility of AI writing tools has forced us to rethink how we can help students improve writing skills both without AI tools, but also alongside AI tools. Teaching students how to write well without the use of AI tools improves critical thinking and learning and can give students the confidence to write without AI. Furthermore, it can teach them to recognise what good writing looks like and this is important when they write alongside AI tools. It improves their ability to effectively judge and make decisions about AI-generated texts. Training students to write alongside AI tools can help them understand these tools’ capabilities and limitations, build confidence in using them, and develop their sense of agency in their use of AI tools.

  Again, the writing pie provides a useful framework for thinking about which aspects of writing you want to help students develop. Here, we give some suggestions for activities working with AI writing tools.

  Of course, writing activities that existed before AI-tools can still teach students a great deal about good writing, and we are currently working on a Language Toolkit to support students in their language and communication. There will be resources addressing a range of language topics, including activities to help develop student writing skills both without and with AI. There will be materials that you can use in your own classes, and also self-study materials for students. You will be able to use these in your courses or adapt them to suit your needs. Please email us at alp.sgw@vu.nl if you would like to be added to that course when it is ready.

  Writing alongside AI writing tools

  • Using an AI tool to edit a text and develop language skills: Ask students to input a draft of a paragraph into an AI writing tool and ask for feedback. This could be an example paragraph or a paragraph that they have written. Guide them in developing a prompt. The University Library e-learning Studying smart with AI gives good tips on making prompts more specific. With language revision it is good to instruct the AI tool to give multiple options and to give explanations of changes rather than simple corrections. Encourage students to be questioning and sceptical about the AI output. You could use this activity to work on accurate grammar and punctuation, a well-connected and coherent text, or style.
  • Comparing AI-generated summaries to student summaries: Ask students to independently summarise a section of text and then compare their work to summaries generated with AI tools. This can help students engage with sources and explain ideas, concepts and theories. Everyone’s summary should be individual because it represents their own understanding and evaluation of the text. Encourage the students to reflect on the results. You could include some guided reflection questions. For example:
    • Did the AI tool summary identify the same points or different points?
    • Did it prioritise the same points?
    • What aspects of critical engagement (evaluation, questioning, interpretation) are outside the scope of this AI-generated summary?
    • How did your learning differ when reading and summarising yourself compared to asking the AI tool to do it?
  • Editing AI-generated text: If your course permits students to use AI tools to help edit texts, you might want to help them understand the problems associated with this process. Your course may even allow students to generate texts with AI, and then you might want to give training in the common features of AI-generated texts. You could ask students to edit or create a short text using an AI tool (2-3 sentences). The students could identify potential problems with the writing, and then edit the text and explain the rationale behind their decisions. You could demonstrate the process first, working with students to identify issues, possible solutions and why the changes might be necessary.
  • Analysing AI-generated text and human-written text: Exploring the differences between human written texts and AI-generated writing can help students become more aware of the features of AI texts and potentially increase their appreciation for their own voice. Present two texts to the students and ask them to read them and identify which they think has been written by a human, and which has been written by an AI. Ask the students to identify the reasons behind their answer. This activity could help students become more aware of voice and style, but also the limitations of AI-generated texts regarding content and organisation.
  • Incorporating reflection on AI tool usage: You can incorporate activities that encourage students to reflect on AI feedback to their texts. FFor example, you could ask them to include a short reflection after a piece of writing. Khojasteh et al. (2025) present an example asking students to discuss the following prompts: Understanding AI's Rationale; Agreement with AI Feedback; Impact on Writing Process; and Areas of Confusion.

• Useful links

  • The VU AI Literacy Companion provides an in-depth guide to students about how they might use AI tools in their studies.
  • The VU didactic tips page has a selection of tips about teaching and AI tools.
  • The University of Sydney: AI for Educators Canvas course provides a lot of tips and suggestions about how educators can incorporate AI tools into their practice. https://educational-innovation.sydney.edu.au/teaching@sydney/what-to-do-about-assessments-if-we-cant-out-design-or-out-run-ai/

• References

  Dreschler, G., Gambrel, A., & Branum, J. (2025). A Case Study in Mindfully Integrating AI Tools into Writing Classes. Journal of Academic Writing, 15(S2), 1–7. https://doi.org/10.18552/joaw.v15iS2.1117

  Evangelista, E. D. L. (2025). Ensuring academic integrity in the age of ChatGPT: Rethinking exam design, assessment strategies, and ethical AI policies in higher education. Contemporary Educational Technology, 17(1). https://doi.org/10.30935/cedtech/15775

  Goulart, L., Matte, M. L., Mendoza, A., Alvarado, L., & Veloso, I. (2024). AI or student writing? Analyzing the situational and linguistic characteristics of undergraduate student writing and AI-generated assignments. Journal of Second Language Writing, 66. https://doi.org/10.1016/j.jslw.2024.101160

  Khojasteh, L., Kafipour, R., Pakdel, F., & Mukundan, J. (2025). Empowering medical students with AI writing co-pilots: design and validation of AI self-assessment toolkit. BMC Medical Education, 25(1). https://doi.org/10.1186/s12909-025-06753-3

  Li, Y., Sha, L., Yan, L., Lin, J., Raković, M., Galbraith, K., Lyons, K., Gašević, D., & Chen, G. (2023). Can large language models write reflectively. Computers and Education: Artificial Intelligence, 4. https://doi.org/10.1016/j.caeai.2023.100140

  Liu, D. (2024, 6 Nov). Multiple Modalities and Generative AI. [Video] YouTube. www.youtube.com/watch?v=FUpoRTyNvVE.

  Perkins, M., Roe, J., Postma, D., McGaughran, J., & Hickerson, D. (2024). Detection of GPT-4 Generated Text in Higher Education: Combining Academic Judgement and Software to Identify Generative AI Tool Misuse. Journal of Academic Ethics, 22(1), 89–113. https://doi.org/10.1007/s10805-023-09492-6

  Savini, C. (2025, July 16). We Can’t Ban AI, but We Can Friction Fix It. Inside Higher Ed. https://www.insidehighered.com/opinion/career-advice/teaching/2025/07/16/we-cant-ban-generative-ai-we-can-friction-fix-it-opinion