Thinking aloud: how I employ AI without losing my voice
I’m someone who processes verbally. I need a sparring partner to clear my thoughts, especially when I’m stuck or circling something complex. I’ve always been this way.
For most of my life, that partner has been my dad. We’ve had a strong intellectual relationship since I was a child. Our long and late conversations taught me how to reason, reframe, and persist with an idea. But now we live on different continents, and he often gently reminds me: “This topic isn’t really my field.”
At work, everyone is busy and then tired. My friends are, too. I’ve learned that asking people to help me untangle academic ideas can become a burden, even when they care. That’s where AI fits in. It doesn’t replace real conversation, but it gives me someone to talk to when I need to think aloud. It’s available, consistent, and responsive. I can ask it to question me rather than give me answers—and it does.
That back-and-forth—me trying to explain, asking for clarification—helps me find the words. Often, thoughts circling for days suddenly shape themselves in minutes, making me shout out loud: “YES!”
A Collaborative Tool, Not a Crutch
I don’t ask for full paragraphs. I ask for structure. For rewording. For a better verb. Sometimes I paste in a rough note and say, “Help me make this sound like me, but in academic English.”
When it works, it sharpens me. When it doesn’t, I revise or delete. It’s more collaboration than automation. Like having a supervisor review a messy draft—not to rewrite it, but to help me see where it’s not quite landing.
And because I have a good idea of how large language models work, I’m intentional about how I use them. I take care to “train” my AI tool in my tone and my voice. That means regularly adjusting what it reflects back to me. I don’t let it drift into generic prose—I nudge it, shape it, and hold it to a standard that sounds like me.
English isn’t foreign to me—but academic writing in English still feels like translation. Not literal translation from Afrikaans, but from something deeper: the way I learn and think. Tone, rhythm, and register still take extra effort. AI helps bridge that gap. It lets me focus on what I want to say instead of getting stuck on how to say it.
(More about this here: https://the-emperors-footnotes.ghost.io/between-tongues-on-afrikaans-english-and-academic-insecurity/)
The Guilt of Getting There Too Fast
Here’s the part I’m still figuring out.
Somewhere in me, there’s a belief that real writing means struggling. That crafting every sentence manually is a rite of passage. So when AI helps me say what I mean faster—and in clearer English—it feels… suspicious.
I notice this most clearly while working on my proposal. I do the thinking. I have the ideas. But turning them into something “academically appropriate” drains me. With AI, I can get there faster. But that speed makes me uneasy. Like I haven’t suffered enough to deserve clarity.
And that unease triggers familiar imposter thoughts:
If I say I used AI, will people think I didn’t really write this?
I have even had that conversation with my AI tool—more than once. I haven’t fully resolved it. But I’m trying to name it, rather than let it keep me from moving forward with my research.
Accountability and Authorship
I’m at the beginning of my Master's. But already, I’m learning that the tools I use don’t define the value of my work—my decisions do.
I decide what to keep. I shape the argument. I care how it reads.
In the end, I make use of AI with human oversight, as my school would phrase it. And that’s the crux, isn’t it? Human oversight. Mine.
I’ll never sign my name to anything I can’t account for—every word, every choice, every implication.
I am the writer.
Note: For this post, too, I was able to refine my English phrasing while preserving the integrity of my ideas with the support of the AI tool, ChatGPT (OpenAI, 2025).
If you’ve ever found yourself second-guessing the tools that help you think—AI or otherwise—I’d love to hear how you’ve navigated that terrain. Please leave a comment!