From rewriting AI slop to building my own workflow: the three ways I write articles now

My personal journey from fighting AI-generated text to finding a system that actually works for every type of article

I've been writing with AI for over two years now, first for my consultant job, then for my blog. And like most writers, I started the same way everyone else did.

But that journey took me somewhere unexpected. I went from endlessly rewriting LLM drafts, to building my own automated pipeline for evergreen content, to finally creating the writing companion I actually needed for the articles that matter most.

Here's what I learned along the way.

Where I started: the LLM-generated article (and why it broke me)

Like everyone in 2024, I jumped on ChatGPT and Claude. The promise was irresistible: give it a brief, get a full draft, polish it up, and publish. Simple, right?

My routine back then:

• Set up a project with context and a summary
• Let the LLM generate a complete draft
• Spend hours rewriting it to sound like me

The problem? I wasn't just editing. I was fighting the AI's voice at every turn. The same tedious corrections, article after article:

• Removing em-dashes — because LLMs overuse them like they're getting paid per dash
• Cutting nonsensical sentences that looked correct but said nothing
• Humanizing generic prose that sounded like it was written by a committee of robots
• Fact-checking hallucinations that I couldn't trust

I was doing the hardest part—making it sound human—after the AI had done the easy part. And I was doing it every single time.

Worse still: I wasn't fast enough. I couldn't cover all the evergreen content I wanted. I couldn't react to news quickly enough. My productivity was bottlenecked by the very tool that was supposed to help me.

Something had to change.

The solution I built: agentic programmatic SEO

Scaling evergreen content without the rewrite grind

I realized I was using the wrong tool for the wrong job. The LLM draft approach works for one-off articles, but it's terrible at scale. Every article required the same manual rewrite process. I was burning hours on content that didn't even need my personal voice.

So I built a pipeline.

How my agentic system works:

• Multiple AI agents work in sequence: one finds topics, another develops angles, a third writes research files, a final one drafts the complete article
• I can set gates to review angles or research before the final draft
• Fully automated when I need volume

What changed: Suddenly I could publish 20 evergreen articles in the time it used to take me to finish one. Technical guides, basic how-tos, news roundups—content that needed to exist but didn't need me specifically.

The catch: These articles aren't personal. They're functional. They inform, they rank, they drive traffic. But my voice? Not in there. And that's fine—because that's not what they're for.

This method gave me back my time. But it also made me realize something important: I was missing a third category entirely.

The missing piece: opinion articles with authenticity

Where my voice actually matters, and Levee was born

Even with my pipeline handling the high-volume stuff, I still had articles that needed me. Opinion pieces. Personal reflections. Articles where my perspective was the whole point.

And I kept falling back to the same broken method: LLM draft, then hours of humanizing. Because none of the existing tools understood the assignment.

What I needed:

• An AI that would help me write, not write for me
• Something that preserved my voice from the first word
• A tool that handled research without breaking my flow
• No generic prose to rewrite, no hallucinations to fact-check

So I built it.

Levee works differently because it's designed for writers who care about voice and control:

• Set your voice first: Paste a sample of your writing, and Levee extracts your style. Every suggestion already sounds like you. No em-dashes to delete. No generic prose to rewrite. It's my voice from the start.

• Write & Tab: When I pause, Levee suggests the next word, sentence, or paragraph in gray. I press Tab to accept, or keep typing to ignore. I'm in complete control—word by word, sentence by sentence.

• Context references: I feed Levee URLs, articles, or notes. It writes with that context in mind—like having a research assistant inside my editor. When I'm writing an opinion piece, I can pull in facts without switching to a browser.

• Fact lookup: I highlight a paragraph and Levee finds relevant facts to strengthen my argument. No more breaking my flow to Google something.

• Instant rewrites: I select any text and ask Levee to shorten, expand, or fix grammar. It does exactly what I ask, in my voice.

What this means for my writing: I finally have a way to write opinion articles that feel authentically mine, without sacrificing productivity. I'm not fighting against generic prose. I'm not spending hours "humanizing" anything. I'm just... writing, with an assistant that makes me faster without making me sound like someone else.

My complete workflow now

Three methods, three purposes. I use each for what it's good at.

The key insight I've learned? There's no single right way to write with AI. The smart writer knows which tool fits which job.

Why I built Levee

I built Levee because I was tired of fighting against AI instead of working with it. I wanted an assistant that would help me write better—not instead of me.

No generic prose to rewrite. No hallucinations to fact-check. No spending hours "humanizing" what an LLM generated. Just my voice, my ideas, and an AI that actually helps me express them better.

I built it for the articles that matter most—the ones where my authenticity is the whole point.

If you're tired of the same struggles I had, try Levee free at levee.theleverage.tech—no card required. I built it for writers like us.