There is a crucial difference between using generative AI as a fully-qualified academic expert and using it as a student. An expert draws upon a deeply embedded understanding of the quality standards, concepts, and methods of their discipline and can quickly identify errors, weaknesses, or omissions. As a student, however, you are still in the process of developing those same standards. In this context, the polished and confident tone of language models can create an illusion of fluency: a sense that content is correct and easy to understand simply because it is well written.

For precisely this reason, the goal is not to use AI as a substitute for the academic judgement you are still developing. Rather, the aim is to use the technology in ways that help you build the disciplinary standards and critical judgement that experts already possess.

What can I do myself?

To avoid the expert trap and challenge the false sense of rapid understanding, you can incorporate the following two simple techniques into your study practice:

Do a '"brain-start": Before you even open your chatbot, spend a couple of minutes writing down your own initial hypotheses, concepts, or informed academic guesses. By consciously retrieving your existing knowledge first, you create a disciplinary reference point in your mind. This makes it less likely that you will accept the model's first response uncritically.

Treat the AI as a student and teach it yourself: Reverse the roles to test whether you have genuinely understood the material. Explain a challenging academic concept to the chatbot in your own words and ask it to act as a curious fellow student who identifies gaps, weaknesses, or misunderstandings in your explanation. Since teaching others is one of the most effective ways of learning, this approach forces you to engage with the material in depth and reveals genuine gaps in your own understanding.

When discussing the use of generative AI, there has been a strong emphasis on prompting as a core skill. It has even acquired the label "prompt engineering". However, this term can create the mistaken impression that a well-written prompt is sufficient in itself. In reality, your true superpower is your deep, subject-specific knowledge that enables you to recognise when AI overlooks important nuances or simply hallucinates facts.

Prompting is a useful tool. However, the value of a prompt not only depends on how precisely it is formulated; it also depends on how critically you examine, challenge, and refine the responses you receive. Developing sound academic and professional judgement is therefore at least as important as developing effective prompting skills.

What can I do myself?

You can use the following technique to counter the authoritative tone of the language models:

Challenge the model's disciplinary perspective: Language models write with a strong, built-in sense of authority, regardless of whether they omit important nuances or simply invent and hallucinate facts. Make a habit of asking follow-up questions such as: "On which theoretical perspective or school of thought is this answer based?" and "Which counter-arguments or disciplinary assumptions have been left out?" Doing so helps break the illusion of the "neutral machine" and reminds you that AI-generated outputs represent only one particular perspective, shaped by the data on which the model was trained.

When you interact with a chatbot, the system is designed to provide a rapid, fluent, and virtually frictionless response. This is a deliberate design feature that may appear quite satisfying at first glance. In the context of  learning, however, it creates a risk of premature coherence: i.e. the AI’s response can appear complete, logical, and persuasive before you have undertaken the necessary intellectual work yourself.

A valid countermeasure could be to introduce reflective friction to your study process. This means deliberately slowing down, asking challenging follow-up questions, testing claims against your course literature, and allowing the effort involved in engaging with the material to generate a deeper understanding of your subject.

What can I do myself?

These are three straightforward strategies that you can apply in your study practice:

Seek contrasts rather than easy answers: Ask the AI to present three different disciplinary perspectives on the same topic side by side. This forces you to actively compare and evaluate the nuances rather than simply accepting the first response.

Use AI to critically review of your drafts: Once you have written a draft of a theoretical discussion or an argument, upload it and ask the model for a critical review. Be specific: "Which methodological weaknesses, disciplinary counterarguments, or unsupported claims stand out in this text?"

Explain your own reasoning: Before accepting a revision, or a new structure suggested by AI, explain to yourself (perhaps even to the model itself) why you choose to give preference to that particular solution and not the others. When forced to justify and verbalise your decisions, you strengthen your metacognitive awareness and sense of academic ownership, both of which you may later be required to demonstrate in an assessment or examination setting. For example: "I am building on the proposed structure for the theory section, but I am changing the methodological approach because the original version overlooks the specific perspective that our course places emphasis on ..."

For a student, critical thinking and the use of generative AI are fundamentally about developing your subject expertise, maintaining academic ownership, and taking responsibility for your own learning processes. Learning, by its very nature, requires cognitive effort, and it is precisely the challenge inherent in that process that leads to lasting a understanding. This is in direct in contrast to the rapid output provided by generative AI.

Once you understand the pitfalls of the technology, and learn how to mitigate them, you can still make productive use of generative AI to e.g.  challenge your thinking, explore topics from multiple angles, and strengthen the knowledge for which you are ultimately accountable.

Critical thinking is not about avoiding AI; it is about retaining intellectual and academic ownership of the learning process. You can view the strategies on this page as practical ways of using AI as a learning partner rather than delegating parts of the thinking process to the technology.