The Voice Interview Method
I have a problem that I suspect is common among academics: I know more than I can articulate. I have expertise and opinions and ideas, but when I sit down to write, they don’t come out in any coherent order. The expansively blank Word document feels impossible to fill. I write a paragraph, delete it, write another paragraph, delete that too. The ideas are in there somewhere, but the structure isn’t.
But lately, I’ve started to enjoy writing again, thanks to AI. I have a new method, which I call the voice interview method, and it’s become one of my most reliable tools for turning messy thinking into structured output.
The Problem
The blank page is a terrible place to organize your thoughts. Writing requires you to do two things at once: generate content and structure it. For complex topics, where you have years of expertise but no clear throughline, there’s too much cognitive load. You end up either writing in circles or staring at the cursor.
Outlining first doesn’t always help either. If your ideas are scattered, an outline is just a structured list of scattered ideas. You need to externalize your thinking before you can organize it.
The Method
Here’s what I do:
Step 1: Turn on voice mode and ask AI to interview you.
I start a conversation with something like: “Interview me about [topic]. Ask me probing questions to help me articulate what I know and think. Push back when I’m vague. Ask follow-ups.”
Step 2: Talk.
This is the part that feels weird at first. You’re just… talking. Out loud. To your phone or computer. About your area of expertise.
But talking is easier than writing for initial externalization. You don’t edit yourself as much. You follow tangents. You say things like “well, actually, the real issue is…” and suddenly you’ve articulated something you’ve been circling around for months.
The interview format helps because it inspires you with questions rather than generate structure from nothing. When the AI asks “why does that matter?” or “can you give an example?” you have to answer. And in answering, you articulate things you hadn’t quite put into words.
Add in that AI isn’t a person, and won’t be offended if you go off script. You can start to answer and then say, “Oh wait, that reminds me of something totally different,” and it will go with the flow. It’s not going to judge a poor choice of words or get impatient if you ramble.
My favorite part of this? I often use this method while multitasking: washing dishes, folding laundry, walking the dog. It feels more natural to talk while doing something else. My brain is more free to roam. And it gives me just a little more downtime at the end of the day. (Just a tip for all you working parents out there!)
Step 3: Ask AI to organize your responses into themes.
Once I’ve talked through the topic—usually 10-20 minutes, sometimes longer—I ask: “Based on what I said, what are the main themes or categories? Group my ideas into buckets.”
This is where the magic happens. AI is good at pattern recognition. It will pull out the throughlines you couldn’t see while you were talking. Sometimes it identifies a theme I didn’t realize I kept returning to. Sometimes it groups two ideas together that I thought were separate but are actually the same point.
I review and adjust. I know my content better than AI does, so if it’s grouped something wrong or missed a key theme, I correct it.
Step 4: Ask AI to arrange the buckets into a logical flow.
Finally: “Put these themes in an order that would make sense for [a blog post / a handbook chapter / a review paper / a lecture].”
The output format matters here. A blog post has different structural needs than a literature review. A handbook chapter needs to build skills sequentially. I tell AI what I’m writing so it can sequence the themes appropriately.
This becomes my working outline.
A few things are happening here that make this more effective than just sitting down to write.
Talking is lower stakes than writing. When I write, I edit myself constantly. When I talk, I don’t. The interview format gives me permission to think out loud without worrying about whether it’s polished.
The interview structure forces articulation. Left to my own devices, I’ll circle around ideas without pinning them down. But when someone asks “what do you mean by that?” I have to answer. Even if that someone is an AI.
Generation and organization are separated. This is the big one. I’m not trying to generate content and structure it at the same time. I generate first (by talking), then organize (with AI’s help). That’s a much more manageable cognitive load.
I retain ownership of the content. The ideas are mine. I said them. AI just helped me see the structure that was already there. This matters for voice—the final piece sounds like me because it started as me talking.
A Note on Voice Mode Specifically
You could do this with text—type your responses instead of speaking them. But I find voice mode meaningfully better for this purpose. Talking is faster, less filtered, and more natural. I say things out loud that I would never type, because typing feels like writing and writing triggers my inner editor. Voice mode keeps me in generation mode longer.
If you haven’t tried voice mode for anything yet, this is a good use case to start with.
Try It
Next time you’re stuck on something you know well but can’t seem to write, try this:
- Open voice mode
- Say: “Interview me about [topic]. Ask probing questions.”
- Talk until you’ve said what you know
- Ask for themes, then logical order
- Use the output as your outline
The ideas are already in your head. Sometimes you just need someone to ask the right questions.