A student submits a perfectly formatted research paper. The citations check out. The argument flows. But can you tell if they actually read the sources?
This is the question keeping many educators up at night. And it's revealing something we may have overlooked for years.
For some time now, I've been seeing the phrase "writing is thinking" used as a defence of traditional writing instruction in response to AI. The argument is usually this: if AI writes for students, students will no longer learn how to think.
I agree with the premise that writing does develop logical, critical, and creative thinking.
But I also think this framing risks missing a deeper issue.
Writing as we know it is changing. And it will continue to change because of AI. Defending past practices without questioning what writing is actually for may be the wrong battle to fight.
So perhaps the more important question is not whether students should write, but what we need writing for in higher education today.
In universities, students, researchers, and lecturers write primarily to communicate research and ideas. But writing does not exist in isolation.
There is no writing without reading.
Which leads to a question that is asked far less often: How do we teach students to read? How much do they actually read? And how critically do they engage with what they read?
Higher education has long emphasised research output and written production. But reading ( slow, critical, comparative, sceptical reading) is often assumed rather than taught.
Perhaps it always was a weak spot. AI has simply made that weakness visible.
Here's what I've noticed in my work. Most of the current debate focuses on whether AI can generate essays, reports, and articles. But I am less concerned about how AI writes and more concerned about how AI reshapes research practices.
AI changes:
1) how students search for information
2) what sources they encounter
3) how quickly they accept answer
4) show often they verify claim
5) show bias is reinforced or hidden
6) how hallucinations pass as knowledge
7)how confirmation bias is amplified
The real risk is not that AI produces text.
The real risk is that it quietly becomes the filter through which knowledge is accessed and trusted.
If students rely on AI to summarise, explain, and select information, then the core academic skill is no longer writing alone. It is epistemic judgment, i.e. deciding what counts as reliable, meaningful, and valid knowledge.
This does not mean we should abandon writing. But it does mean we should rethink its role.
Instead of defending writing as thinking in isolation, we may need to emphasise:
- critical reading
- source evaluation
- bias detection
- research literacy
- reflective use of AI
- and the ability to question generated knowledge
In that sense, AI has not destroyed academic skills. It has exposed where they were already fragile.
Whether you're teaching, hiring, or learning, these questions can help maintain epistemic rigour:
- Can you trace the claim back to its original source? Not just a citation, but the actual primary research or document.
- What perspective or bias might be embedded in how this information was selected or summarised? AI tools have training biases. So do we.
- What would change your mind about this conclusion? If you can't answer this, you may be accepting information uncritically.
The debate, I think, is slowly shifting. It is no longer just about whether AI can write. It is about how AI transforms how we know, how we research, and how we decide what to trust.
And that may be the more urgent educational challenge.
