The Free AI Blog Engine I Built, Then Gave Away

Most tools that promise to write your blog skip the part that actually gets a post ranked. They draft fast and leave you the real work: checking the topic isn’t competing with a page you already have, structuring the piece so AI Overviews and featured snippets can quote it, placing internal links that make sense, setting the SEO fields, and rewriting the bits that read like a machine. So I built one that does the whole job. First for a niche marketplace I run on the side, and now it’s free on GitHub for anyone using Claude Cowork.

You can grab it at the end of this post. First, why it exists and what building it taught me.

The problem I was actually solving

Literally every project these days needs a steady stream of content, and writing each post by hand is slow. Every AI writer I tried from Jasper, Surfer to ChatGPT and Claude, had the same gap. It handed me a clean draft and left the editorial work to me: confirming the topic wasn’t cannibalising a page I already ranked with, finding the right internal links, shaping the piece so it could earn a snippet, and rewriting anything that sounded generated.

That last part matters more than people like to admit. A draft that reads as AI doesn’t build the trust that makes someone act on it. I wanted a system that treated the draft as one step in a bigger editorial process, not the finish line.

What it actually does

The engine runs the full job end to end and hands me a finished article to review. It:

• Researches the topic live and writes from real, dated facts, so nothing is invented.
• Follows a set of article types (guide, comparison, listicle, FAQ, opinion, review, pillar), each with the structure that fits its search intent.
• Checks every topic against a keyword map, so two of my own posts never fight for the same query.
• Places real internal links, and keeps external links nofollow and opening in a new tab.
• Writes the SEO title, meta description, and slug to current best practice, and structures the piece answer-first for AI search.
• Runs a humanising pass so the writing sounds like a person.
• Builds a branded featured image.
• Delivers the result either straight into WordPress as a draft, or as a Markdown and Word file if you’d rather nothing auto-published.

It also remembers what to write next in a simple queue, so it keeps going without me holding the plan in my head.

What building it actually taught me

There were plenty of technical gotchas along the way, and I wrote every one into the repo so nobody else loses the nights I did. But the lessons worth carrying into the next project have nothing to do with the code.

The one that caught me off guard is that I didn’t really understand my own content process until I tried to hand it to a machine. For years the editorial checks lived in my head, and I ran them unevenly, more on the weeks I had time. Turning them into steps a tool could follow forced me to say out loud what “good” actually means, which check matters when, and why. That made me sharper at the manual version too.

Consistency also turned out to matter more than speed. I built this to save time, but the real payoff was that the boring, skippable steps now run the same way every time, including the weeks I’m stretched thin and would otherwise cut a corner. A process that only holds up when you’re at your best isn’t much of a process.

It also settled a question I go back and forth on, whether to build your own tools or just buy something close. Building it kept me next to the problem, so I learned exactly where content work breaks and what a good fix looks like. That understanding carries into everything else I do, whether or not I open the tool again.

Hope this inspires someone to create the next best tool 🙂

Why I gave it away

This is the second free tool I’ve published this year, after the marketing pre-launch campaign audit skill. And it is deliberate. Building something useful and giving it away teaches me more than sitting on it would, and it puts real work in front of the people I want to learn from. The engine is fully generic now: no trace of my own site, a guided setup that asks about your niche, your voice, and your brand, and two output modes so it fits whether or not you use WordPress.

There’s one thing I keep relearning through all of this. Automating the draft buys you speed, but the real leverage in content sits one step upstream, in the editorial system that decides what to write and how it gets built. Automate the writing alone and you just get faster at producing average posts. Encode the editorial thinking, and the posts themselves get better.

If your role involves writing articles and you use Cowork, this should save you real hours. And if it does, I’d like to hear what it caught, or what it got wrong.

The full blog engine, with install instructions, is free on GitHub. Grab it using the button below.