The Interview Is the Secret: How I Use AI to Write Articles That Actually Sound Like Me
The process behind these posts — and why my mom still can't tell.
My mom has read all of my books. Every single one. She knows my voice the way only a person who’s been reading your words for years can — the rhythms, the opinions, the little declarative punches I throw when I really mean something. So when she told me she couldn’t tell that my Substack articles were written with AI, I knew I was doing something right.
(Hi, Mom.)
I get asked about this a lot. “Do you write these Substack articles with AI? It doesn’t seem like it.”
I always say yes — yes, I do. And most people aren’t doing it the way I do it. That difference is everything.
The Question Everyone Is Actually Asking
When people ask if I use AI and then look surprised when I say yes, what they’re really asking is: how do you make it sound like you? Because they’ve tried AI writing tools. They’ve gotten that smooth, helpful, relentlessly enthusiastic output that sounds like... no one in particular. Competent. Readable. Completely devoid of personality.
That’s not a tool problem. That’s a process problem. And the process is fixable.
Step One: Find the Topic Together
Sometimes I come to a session with a topic already in mind — something I’ve been thinking about, something that came up in the Future Fiction Academy, a conversation I keep having in the Facebook group. Other times I’ll ask for ideas, browse through what the AI surfaces, and wait for one to land. I know it’s right when I feel a little spark of yes, I have opinions about that.
Once I have the topic, I give a quick gut-feel brief. Not a formal outline — just a sentence or two about why I like it and what my overall instinct is. That initial reaction is the seed of everything that comes after. It tells the AI where I’m starting from emotionally and intellectually, which shapes the questions it’s going to ask me.
Step Two: The Interview — This Is the Whole Thing
Here’s where most people check out of the process too early. They give AI a two-or three-sentence prompt and wait for a draft. And they get average output, because they put in average input. Of course it doesn’t sound like them. They haven’t actually put themselves into it.
What I do instead is let the AI interview me.
After I give my initial take, I get asked targeted questions — about my opinions on different angles of the topic, about my personal experience, about the spicy take I might be sitting on, about what I want readers to walk away feeling. And then I answer. Honestly. Sometimes I word vomit. Sometimes I use talk-to-text so I can pace around and just rant into my phone while the words tumble out. I don’t edit myself. I don’t try to sound polished. I say what I actually think, in the way I actually say things.
And this works because AI is extraordinarily good at mimicking. If you give it your voice — your actual opinions, your rhythm, the way you’re declarative when you mean business — it has real material to work with. I’m a Capricorn. I don’t hedge; I state. That comes through in how I answer questions, and it comes through in the draft that follows.
The interview also does something else for me personally: it draws out opinions I didn’t know I’d fully formed yet. My top Clifton Strength is Deliberative — I need to think things through from every angle before I feel confident in my opinion. A good Q&A session is structured introspection. It keeps me focused on the goal when my brain wants to wander down every tangential hallway, and it consistently pulls out the most honest, specific version of what I believe about something.
Step Three: The Draft
Once the interview is done — and sometimes there are a few follow-up questions to nail down a specific point — the AI writes the first draft. It’s not just synthesizing facts. It’s reading for voice, for structure, for the shape of my argument. It knows my previous articles. It has a style guide. It has this conversation. All of that context goes into producing something that already sounds reasonably close to me before I’ve touched it.
This is the heavy lifting I’m happy to hand off. Generating the bones of a 1,200-word piece from a conversation is exactly the kind of task AI does well. The thinking, the opinions, the specific anecdotes — those came from me. The assembly is the AI’s job.
Step Four: The Edit — The Part That’s Always Mine
Here is my one non-negotiable: every word gets read. Every sentence gets validated. I do not generate and post. Ever.
Reading through the draft, I’m listening for anything that doesn’t sound like me. Sometimes it’s a phrase I’d never use. Sometimes it’s a structural move I don’t love. And sometimes it’s what I call AI-isms — the little tics that reveal the seams. My personal most-wanted list:
“It’s not X. It’s Y.” — This construction is everywhere in AI writing. It sounds clever in a way that I don’t actually sound clever. Out.
“Here’s what I learned:” — The warm-up statement followed by a colon, setting up a list or a paragraph. I don’t talk like that. Out. (In fact, I booted one from this very article! Which makes me laugh since the AI knows I don’t like it.)
Anything that feels like a conclusion before the conclusion. AI loves to summarize mid-article. I don’t.
After I’ve made my edits — sometimes light, sometimes more significant — the piece goes to Substack for formatting. But the editing step is where I’m most fully myself in this process. The buck stops with me. I am always the human in the loop. The AI does the heavy lifting, yes, but my input, my final say, my name on the byline — those belong to me so I make sure I get it right.
Why This Works (And Who It’s For)
I want to be honest about what this process is and what it isn’t. It isn’t magic. It isn’t a shortcut to skip the thinking. It’s actually a very structured way of forcing the thinking to happen — and then handing the transcription and assembly to something faster than I am.
The authors who benefit most from this approach are the ones who have a strong internal voice and specific opinions, but who struggle to get those things out of their heads and onto the page in a way that feels coherent. If you know what you think but the blank page is the enemy, this process is for you. The interview replaces the blank page with a conversation. And most of us are much better in conversation than we are staring down a cursor.
If you’ve been handing AI a prompt and getting output that sounds like nobody, try handing it a conversation instead. Tell it your gut reaction. Let it push back with questions. Pace around if you have to. Get your actual voice into the process before any draft gets written.
That’s the secret. It was never about the tool. It’s always been about YOU.
Have you tried an interview-style process with AI, or are you still starting with the two-sentence prompt? I’d genuinely love to know what’s working — or not working — for you. Tell me in the comments.



I left one AI-ism in this article because it made me laugh (again, the AI knows I don't like this thing done too often). But it's actually GOOD writing when done well. Can you spot it?
Hi, Steph. xoxo Mom