This post grew out of a presentation I gave for Turnitin in May 2026, where I was invited to speak as an expert presenter on AI in higher education. A lot of what’s here started as voice notes which feels appropriate given the subject matter.

 

Most of the conversation about AI in education circles around what AI can produce. Can it write an essay? Can it answer exam questions? Can students use it to complete assessments without actually learning anything?
These are worth asking. But I think they’re the wrong starting point.

 

The question I keep coming back to is: how can AI support learning without replacing it?

 

I’ve been using AI pretty extensively over the last few years in my teaching, my research, and just in everyday life. Like most people, I started by using it in fairly obvious ways. Ask a question, get an answer, check whether the answer seems right. But I got increasingly uncomfortable with that model, and it took me a while to figure out why.
The problem is that when AI produces the answer first, you’re placed in the position of evaluating something without necessarily having the knowledge to judge it properly. A correct answer isn’t always a good answer. It can be factually accurate and still be incomplete, superficial, or missing something important, and if you don’t already know the topic, you might not notice what’s absent.

 

So I changed how I use it.

 

Rather than going to AI for information, I use it to process information I’ve already encountered through sources I trust. I read the paper myself. I look at the evidence myself. Then I use AI to help me work with those ideas — asking questions, testing my understanding, exploring connections. The learning stays with me. The AI plays a supporting role.

 

The thing that’s made the biggest difference to that process is voice input.
We think far more than we can type. That sounds obvious when you say it, but the implications are bigger than they first appear.
When we type, we filter. We shorten things. We simplify. All the half-formed thoughts, the tangents, the uncertainties, the connections we haven’t quite articulated yet – they get lost because capturing them is too much effort. By the time your fingers catch up with your brain, you’ve already edited yourself.
Voice input changes that. When I talk to an AI tool rather than type at it, I can ramble a bit. I can explain what I understand and what I’m still not sure about. I can circle back. I can think out loud which, as it turns out, is often when I do my best thinking.

 

People talk a lot about prompt engineering. In my experience, context engineering is far more useful. The more context you can give, the more useful the interaction becomes. And voice makes it much easier to provide that context, because you’re not fighting the friction of the keyboard or the pen.

 

This matters especially for learning, because uncertainty isn’t a problem to be eliminated, it’s usually where learning starts. When learners can explain their thinking, including the bits they’re confused about, that creates the conditions for reflection and feedback. Voice input lowers the barrier to doing that.
It also fits with what learning research actually tells us. Deep learning isn’t just exposure to information; it involves active processing, self-explanation, connecting new ideas to what you already know, making meaning. Voice-supported AI can genuinely help with that, in a way that “generate me an essay” really can’t.
When I’m reading academic papers, for example, I’ll often talk through my thinking with AI as I go. Why does this paper matter? How does it connect to other things I’ve read? Where am I uncertain? The AI helps me clarify and identify gaps but it’s not doing the reading for me. It’s helping me think about what I’ve read.
The same goes for students. AI can help them organise their thoughts, reflect on their learning, generate questions, test their understanding. What it can’t do is the actual learning – the judgement, the interpretation, the meaning-making. That bit is still theirs to do.

 

The promise of AI in education isn’t that it eliminates effort. It’s that it can reduce effort spent on routine tasks and make more room for the cognitive work that actually matters.
Used well, it’s not an answer generator. It’s a thinking partner.
And voice input might be what makes that partnership actually work by reducing the friction between thinking and expression, so that more of our thinking makes it into the conversation in the first place.