Alright DMC crew ...
Bit of a confession: I didn’t mean to build an AI writing system. It just sort of... happened after too many late nights trying to make GPT stop sounding like a bored intern from 2016.
You know that feeling when you almost get the output you want, but it suddenly forgets what it’s writing about halfway through? Yeah, that was me. After about 40 failed runs and a lot of 'zen' breathing trying not to lose my rag, I finally cracked a repeatable setup that writes full SEO-ready articles while sounding (mostly) human.
Thought I’d share it here since a couple of people here have asked how I’m scaling parasite content so fast. This framework works a treat for me. Tweak it however you like, but the scaffolding is solid.
Let’s crack on and just to say, for reference, this is based heavily around Andy's article writing prompt here. I've modified it over time but that prompt will give you a head start if your plan is to recreate your own article writing system based on the framework below.
1️⃣ Start With a Real Discussion
Don’t just dream up topics. I start with a Reddit thread, Quora post, YouTube transcript or comment section that’s already got people chatting.
Example: “SEOs noticing Google rewriting title tags again after the latest update.”
That thread gives you user intent in the wild - what people actually care about - and it helps shape both the keywords and emotional hooks later.
2️⃣ Use GPT Like an SEO Consultant, Not a Writer
Before writing anything, I tell GPT to act like a 20-year SEO veteran (bit of white-hat, bit of grey-hat).
The point is to make it think strategically, not mechanically.
Then I have it:
-
Find 20 high-traffic keywords linked to the topic.
-
Rewrite those keywords into natural sentence-friendly phrases (so you can drop them into text without it sounding forced).
You end up with two lists: raw SEO targets and refined LSI keywords.
3️⃣ Lock in the Workflow
This where Andy's mammoth prompt comes in ;) The secret sauce is the “If → Then” instruction.
I tell GPT:
“If you understand, say ‘Provide me with your keyword.’”
That single line keeps everything modular. Once it replies, I feed it my topic, and then I control the process step-by-step by telling it which Section to run next (A through F).
Each section handles a different part of article creation.
4️⃣ The Six-Section Framework
Here’s what each bit does (this is the logic, not the word-for-word prompt):
Section A - The Outline
GPT builds an outline with SEO-friendly headings that naturally include your refined keywords.
You can add subpoints if needed. This gives you a clean structure that hits every user intent.
When it’s done, you tell it to move to the next section.
Section B - The Understanding Check
GPT summarises its own outline in plain English to prove it’s actually understood what it’s doing.
Think of it like a safety check - this stops the AI from wandering off into waffle.
Section C - The Introduction
This writes a personal, emotional intro (about 150 words).
It starts with your main keyword in the first line, uses a relatable example, and acknowledges how readers feel about the topic.
Banned phrases (like “game-changer”) are stripped out by default in this stage.
Section D - First Main Section
GPT writes the first body section in a chatty, human tone.
It uses examples and keeps things light without losing accuracy.
Each paragraph is short (roughly 170 characters), so it reads nicely on mobile.
Section E - Remaining Sections
It repeats that tone for the other headings - same style, no repetition, clean transitions.
After each one, GPT tells me how many are left so I can decide whether to carry on or skip ahead.
Section F - The Conclusion
It wraps it all up with a catchy title and a call to action, written in first person.
No “Congratulations!” nonsense - just a friendly close and a reminder to visit the main site or share the post.
It ends simply with: “All done :)”
5️⃣ Keyword Integration & Naturalisation
Once the draft’s complete, I run a “semantic keyword coverage” check.
GPT compares the text against the refined LSI list and spots any missing or weakly used phrases.
If something’s missing, it suggests where it fits naturally (or adds a new section title if needed).
6️⃣ The Final Rewrite Pass
This is the polish stage. I tell GPT to:
-
Swap out any banned phrases.
-
Insert missing LSI terms smoothly, no stuffing.
-
Check flow and rhythm for natural readability.
-
Add humour or cheeky lines where they fit.
-
Replace em-dashes with normal punctuation (hyphens, commas, etc).
-
Bold key phrases and keywords for scanning.
-
Keep paragraphs short and lively.
The end result reads like a proper human post, not AI fluff - and every keyword’s covered for SEO.
7️⃣ The Finishing Touches
Once the article’s done, I:
-
Generate 5 Medium-friendly tags. (works for other platforms too)
-
Write a LinkedIn caption that teases the post. (or wherever you're sharing)
-
Keep it short and conversational, something like:
“Google’s at it again - title tags getting rewritten left, right and centre. Here’s why your CTR might be tanking”
8️⃣ Optional Add-Ons
You can even plug this into a Google Sheet + Python script to run automatically for new Reddit or Twitter topics.
As long as the Section logic stays intact, you can loop through it endlessly and it still works.
(I still need to finish that workflow!)
9️⃣ Why It Works
Because it’s process-driven, not prompt-driven.
Most people chuck GPT one massive prompt and hope for the best.
This forces structure, review, and small logical steps - like an actual content workflow, not a magic trick.
🔟 TL;DR
I basically turned GPT into a content assistant that behaves like an SEO strategist, not a copy-paste machine.
Each Section does one job, each pass tightens the quality, and the final output’s good enough to post on Medium, Reddit, or a parasite page without heavy editing.
If you rebuild this in your own way, you’ll get the same control, consistency, and natural tone.
It’s a bit like the old-school article spinner days - except this time it actually sounds decent 😎
Hope this helps someone step it up a gear!
Cheers, Rohan 🍻