Using AI as a revision tool
Artificial intelligence tools such as ChatGPT are increasingly part of students’ everyday lives. I’ve discussed the question of whether AI is useful as a revision tool with many of my students and today, I’d like to share my thoughts.
AI use is not a requirement for academic success
It is important to be clear from the outset that students do not need to use AI tools in order to perform well in exams. For decades, students have achieved excellent outcomes using traditional revision methods such as textbooks, practice questions, and teacher feedback.
With the increasing popularisation of AI, there is a risk that students will feel that they can’t keep up without the use of these tools and so end up relying on them too heavily. AI use has several associated drawbacks, which I won’t go into detail about here, and some students may prefer not to use it at all.
The purpose of this blog post is not to encourage or discourage the use of AI, but rather to teach students already using it how they can use it in a way that doesn’t hinder their learning.
Struggle plays a vital role in learning
Cognitive science research consistently shows that learning is strengthened when we try to retrieve information, test ideas, and confront misconceptions. When students encounter new material, the process of attempting to reason through a problem, making mistakes, and correcting them helps to build durable understanding.
If students turn to ChatGPT as soon as they encounter a problem they don’t instantly know how to solve, even if they go on to understand the reasoning behind the problem, they miss out on that vital learning process of struggling. This pattern can also create an illusion of understanding, whereby students recognise explanations when they see them but struggle to generate those explanations when unsupported.
In my experience, when students over-rely on AI tools to answer all of their questions straight away, the lack of depth in their understanding becomes immediately obvious in an exam setting, since they don’t have the developed resilience and toolkit to work through unfamiliar problems.
Using AI for practice generation
If students are using AI to revise, they should use it to generate practice and enhance understanding, rather than to replace their thinking.
For example, students can use AI to:
Create essay prompts
Produce flashcards for factual recall
Generate practice questions
One particularly effective approach is to provide AI tools with specific exam questions or sections of the specification and ask for targeted practice questions. This allows students to increase the quantity of relevant practice while remaining actively engaged in problem-solving. Crucially, students should attempt these questions independently before seeking explanations.
If, once the student has given the question a go, they are still struggling to find a way through, they could ask the AI tool for a hint. If they’re doing independent practice and have access to mark schemes, they could provide the AI tool with the question and mark scheme and ask it to explain the answer to them in more detail.
Think coach, not shortcut
A helpful way to conceptualise AI is as a coach rather than a shortcut. A coach sets challenges, prompts reflection, and provides feedback, but does not remove the effort required from the learner.
Used in this way, AI can support revision by:
Asking students probing questions about their reasoning
Offering alternative explanations after an attempt has been made
Highlighting gaps in understanding without immediately resolving them
Final thoughts
The impact of AI tools on revision depends entirely on how they are used. When students rely on AI to bypass difficulty, the learning process suffers. When they use it to practice, test understanding, and reflect on mistakes, AI can become a useful, though entirely optional, addition to exam preparation.
The most important factors in academic success remain unchanged: sustained effort, deliberate practice, and the willingness to engage with challenging material. AI may support that process, but it cannot, and should not, replace it.
