I Published 30 Articles in Less Than 30 Days. Here's What I Learned.
Thirty articles in under thirty days. I actually did it.
And I learned a lot in the process. About AI, about my own voice, and about what it actually takes to produce content at scale without losing yourself in the output.
If you’ve been curious about using AI to write, or if you’ve tried and couldn’t get it quite right or to sound like you, here’s what I want to share with you.
Claude Changed Everything
Like everyone else, I was using ChatGPT for everything, including writing. But it just never sounded right, especially not for writing long-form content. When I started working with Claude, everything shifted. The quality of the output, the way it could hold context, and the way it responded to direction. It was so much better than anything I’d experienced before, and it’s really what made publishing at this pace possible.
I’m not saying ChatGPT can’t work for people. But for me, for my voice and the kind of writing I wanted to produce, Claude was what finally made it work.
The Difference Between AI-Assisted and AI-Generated
A lot of people misunderstand what it means to write with AI.
What I’m doing is AI-assisted writing. I am still the author. I’m directing the ideas, shaping the structure, reading every line, and making sure what goes out sounds like me and reflects what I actually believe. Claude is a tool in that process, and I remain responsible for the final product.
AI-generated content is something different. That’s when you push a button, take whatever comes out, and publish it without really touching it. I don’t work that way, and honestly, I don’t think that approach produces anything worth reading.
The distinction matters. If you’re using AI to write, you have to decide which one you’re doing. I chose to stay the author. The tool accelerates my process, but it doesn’t replace my voice or my responsibility for what I publish.
Guardrails Became Non-Negotiable
AI-written content has recognizable patterns and phrases that make it obvious it wasn’t written by a human. There are patterns, phrases, and structures that creep in and make your writing feel generic even when the ideas are good.
Before I started writing on Substack, I already had a set of writing guardrails that I would use in ChatGPT or whatever AI software I was working in. I thought they were comprehensive. I thought I had it covered.
I was wrong.
Once I started writing actual articles in Claude, I realized very quickly that my guardrails were not sufficient. They weren’t complete enough for the kind of writing I wanted to produce.
So initially each article took forty-five minutes to an hour to draft, edit, and publish. It was taking FOREVER. Most of that time was spent reading through articles line by line, identifying all the AI phrasing and patterns, and building out an extended guardrail document as I went.
But that’s how I ended up with a guardrails document that actually works. I shared it with my paid subscribers yesterday. If you’re serious about writing with AI and you want to see exactly what I use to keep my content sounding like me, you can subscribe and get access to it.
The Three-Pass Review
Having guardrails isn’t enough if you don’t actually use them. I manually enter my guardrail list three times and have Claude check a drafted article against them before I ever even review it for content and context.
Three passes on every article. That’s how I catch the patterns I don’t want slipping through.
It sounds like extra work, and it is. But that’s how you end up with content that actually sounds like you wrote it.
Claude and I Had to Learn Each Other
There was a learning curve on both sides.
Claude had to learn me. How I like things structured, what my voice sounds like, and what patterns I want to avoid. And I had to learn how I wanted to write. I’ve done long-form posts on Facebook before, but I’ve never written an official blog like this. I had to figure out my format, my rhythm, and what kind of articles I actually wanted to create.
That mutual learning process took time. But as Claude learned my preferences and I refined my guardrails and got clearer on my own style, the whole thing started to flow. What used to take an hour now takes about twenty minutes for most articles.
Some articles still take longer. I wrote a piece the other day that was more biblically focused, and that one took me two hours because I basically wrote the whole thing myself through Claude. Some writing requires more of you, and the tool doesn’t eliminate that.
You Can’t Write About What You Don’t Know
One of the clearest lessons from this process is that you cannot write with AI about something you know nothing about. It just won’t work.
If you don’t know the subject deeply, you won’t catch when AI is being too generic or when it’s actually wrong, and it is wrong sometimes. You’ll end up publishing content that sounds like everyone else’s because you don’t have the knowledge to push back and make it specific.
AI works best when it’s amplifying a voice that already has something to say. You bring the expertise, the perspective, and the lived experience, and AI helps you get it out faster and cleaner. But if you don’t have anything real to bring, AI will just produce noise.
The Fatigue Was Real
I’m not going to pretend this was effortless. Writing thirty articles in thirty days is a lot. There were moments, usually late in the day when I realized I hadn’t written yet, where it felt like a chore. In fact, I’m editing this article at 7pm and have to get it published today. Didn’t really feel like it but I’m committed to an article a day for 30 days. So sometimes it feels like a chore. There have been days, like today, where I was tired and the last thing I wanted to do was sit down and produce another piece of content.
But every single time I started, I got energized. I’d get into the ideas, get excited about what I was writing, and by the end I’d feel good about what I’d created.
Still, I was getting fatigued. So moving forward, I’m scaling back. Three articles a week for free subscribers, two articles a week for paid subscribers. The paid posts are going to be much more in-depth, behind-the-scenes content about how I’m actually building my business, the tools I’m using, and practical application for making money online with AI.
That’s five posts a week instead of nine, which is sustainable and will actually improve the quality.
Why I Never Wanted to Quit
I never had a moment where I wanted to stop. Even when I was tired, even when it felt like a lot, I could see that I was building something I’d be proud of in hindsight.
Part of that is just discipline. I’m a fairly disciplined person by nature, and I know that’s not everyone’s wiring. But part of it was also the strategy behind what I was doing.
I’m connecting my Substack to Pinterest. The pins I create are indexed, and that content doesn’t disappear into an algorithm the way posts do on Facebook or Instagram. This work compounds. Years from now, people will still be finding these articles through Pinterest. The effort I put in this month will keep paying off long after the month is over.
AI is moving fast and things are shifting constantly, but the foundation I’m building now will keep bringing people to what I’m teaching. That made it feel worth it, even on the hard days.
I’m Learning While I Teach
One of the things I love most about this process is that I’m learning at the same time I’m teaching.
There’s a lot I already know about AI. But as I put my ideas into Claude, as I tell it what I want to write about, it makes suggestions. It surfaces angles I hadn’t considered and sends me down paths I wouldn’t have explored on my own.
That’s always the goal for me. I always want to be learning more. And this process is delivering that in ways I didn’t expect.
The Voice-to-Text Hack
I picked up a trick from a mentor that I absolutely love.
When I have a thought I want to give Claude, something I want included in an article, I don’t type it out. I open ChatGPT, use the voice record button, and just talk. I say what I’m thinking in my own words, the way I’d actually say it. Then ChatGPT transcribes it and I paste it over into Claude.
That’s actually how I wrote the notes for this article. I just talked into the microphone and let it capture my thoughts.
What this does is give Claude my actual voice from the start. My phrasing, my rhythm, and my way of saying things. The input is already conversational, so the output comes back closer to how I really sound.
It’s a small shift but it makes a real difference.
If You’re Thinking About Starting
If you’ve been considering starting a Substack or a blog, I’d challenge you to try thirty articles in thirty days.
It sounds like a lot, and it is. But you will learn so much about yourself. About how you think, how you write, and what you actually have to say. You’ll figure out your voice faster than you would writing once a week, and you’ll build something real instead of just thinking about it.
And if you use AI to help you do it, you’ll learn exactly where the tool helps and where you still have to show up yourself.
I’m so glad I did this. It taught me more than I expected, and I’m genuinely proud of what I built.
If you want to see the exact guardrails I use to keep my writing sounding like me, I shared the full document with paid subscribers. Subscribe and you’ll get immediate access, plus everything else I’m building here as I learn and teach my way through the AI shift.