7 AI-Written Articles Got Me 416K Views (copy my exact system)
Research, outline, draft, edit, headline variations, thumbnail designs. By the time I hit publish, I was exhausted.
I spent 3 hours writing my first X article.
Research, outline, draft, edit, headline variations, thumbnail designs. By the time I hit publish, I was exhausted.
Then I realized: I'm literally building AI automation tools. Why am I doing this manually?
So I built a skill that writes entire X articles for me. Topic research, body copy, headlines, thumbnails - all automated.
In the last 2 weeks, it's published 7 articles that generated 416,700 views, 3,743 likes, and 10,714 bookmarks.
Here's exactly how it works (and how you can steal it).
PS: I'm launching a community where I teach this entire system (brand voice files, skill building, agent setup, and more).
If you want to build automations like this for your business, join the waitlist: https://return-my-time.kit.com/1bd2720397
-----------------------------------------
The 7-Step Automated Workflow
1. Topic Research (Automated)
I don't pick topics anymore. The skill uses the last30days tool to scrape Reddit and X for trending conversations in my niche.
It looks for:
- High engagement threads (100+ upvotes, 50+ replies)
- Repeat questions (same problem asked 3+ times)
- Tactical discussions (people sharing specific tools/workflows)
Output: Top 3 topic suggestions with engagement data.
If I don't provide a topic, it picks the best one automatically and proceeds to step 2.
2. Write the Article Body (700-1000 Words)
The skill writes in my brand voice by referencing my Brand Voice skill with:
- My writing samples
- Tone preferences (direct, no-fluff, tactical)
- Sentence structure patterns
- Common phrases I use
Structure it follows:
- Numbered tactical list (5-10 items)
- Each item: Title + Explanation + Specific example
- Short paragraphs (1-4 sentences max)
- Real numbers, pricing, tools mentioned
Time saved: 90 minutes → 5 minutes.
3. Save to Google Drive
Article gets uploaded as a Google Doc to my Drive automatically:
gog drive upload "article.docx" \
--name "X Article: [Topic]" \
--account [email protected]Why Google Docs? Easy collaboration. I can review/edit from my phone before publishing.
4. Write Intro + CTA
The skill prepends a 50-100 word intro:
- Hook: Bold claim or curiosity gap
- Context: Why this matters now
- Promise: What you'll get from reading
- CTA: Lead magnet link
Default CTA: AI Action Plan quiz (my main lead magnet).
Alternative CTAs: Skill Builder GPT, Agent Blueprint GPT (depending on article topic).
5. Generate 7 Headline Options
This is where most people fail. One weak headline kills a great article.
The skill generates 7 variations using proven patterns:
Patterns that work:
- Number + Curiosity: "20 Setups That Made Me Say 'Wait, It Can Do That?'"
- Ultimate + Secret: "The Ultimate Guide (the 3 features nobody told you about)"
- Time + Steps: "how I built X in 2 hours (5 steps)"
- Money + Urgency: "5 Workflows Saving Me 10+ Hours/Week Right Now"
I pick the best one (usually takes 30 seconds).
6. Create 3 Thumbnail Designs
Specs: 5:2 aspect ratio (1200×480px) - required for X articles.
Process:
- Create 3 HTML files with different designs
- Use headless Chrome to render at exact dimensions
- Verify with sips -g pixelWidth -g pixelHeight (learned this the hard way)
3 Variations:
- Stacked Cards: Light background, bold headline left, 3-5 cards right
- Clean Text: Dark/light solid color, large typography, minimal graphics
- Icon Grid: 2×3 grid of icons, headline overlay
Time saved: 45 minutes → 8 minutes.
7. Deliver Everything to Discord
One message with:
- Google Doc link
- Word count
- 7 headline options
- 3 thumbnail PNG files
Ready to publish in under 15 minutes total.
-----------------------------------------
Real Results (7 Articles)
Here's what the automated workflow actually produced:
- Article 1
Engagement: 28,344 views | 308 likes | 27 reposts | 767 bookmarks - Article 2
Engagement: 140,504 views | 1,344 likes | 133 reposts | 3,760 bookmarks - Article 3
Engagement: 7,898 views | 55 likes | 6 reposts | 153 bookmarks - Article 4
Engagement: 37,967 views | 246 likes | 20 reposts | 630 bookmarks - Article 5
Engagement: 16,449 views | 157 likes | 16 reposts | 412 bookmarks - Article 6
Engagement: 95,004 views | 812 likes | 72 reposts | 2,469 bookmarks - Article 7
Engagement: 90,729 views | 821 likes | 86 reposts | 2,523 bookmarks
TOTAL AGGREGATE:
- 416,700+ views
- 3,743 likes
- 10,714 bookmarks
- Average time per article: 12-15 minutes
The Tech Stack (What You Need)
1. OpenClaw - The agent platform that runs the workflow
2. Claude API - Powers the writing (I use Sonnet 4.5)
3. Google Drive API - Saves articles to Drive
4. Last30Days skill - Handle topic research (uses Reddit/X APIs)
5. Headless Chrome - Renders thumbnails
Total cost: $20-80/month depending on article volume and API usage.
What You Can Steal
You don't need to build the exact workflow. Here's what's replicable:
1. Brand Voice Profile Create a /brand-voice/ folder with:
- 3-5 writing samples (your best content)
- Tone guide (formal/casual, technical/simple)
- Common phrases you use
- Sentence structure preferences
Your agent references this file when writing.
2. Headline Pattern Library Save proven headline formulas you see working:
- Track your top 5 performing articles
- Extract the patterns (number + curiosity, time + steps, etc.)
- Make a reference file your agent uses
3. Thumbnail Templates Create 2-3 HTML thumbnail templates you like. Change text/colors per article, render with headless Chrome.
4. Research Prompts Use last30days or similar tools to find trending topics:
- What are people asking repeatedly?
- What posts got 100+ upvotes this week?
- What problems need tactical solutions?
The Pattern You Should Notice
Every step is deterministic.
There's no "creative magic" here. It's:
- Research trending topics (data)
- Write in my voice (reference file)
- Use proven headline patterns (pattern library)
- Generate thumbnails (HTML templates)
That's why it works reliably.
I'm not asking AI to "be creative." I'm asking it to follow a system.
What I'd Do Differently
1. Add engagement prediction Before publishing, analyze the headline/intro against past top performers. Predict likelihood of 50K+ views.
2. A/B test headlines Publish 2 headline variations, track engagement for 2 hours, promote the winner.
3. Repurpose automatically Turn each article into:
- LinkedIn post (pull 3 key points)
- Newsletter segment
- YouTube script outline
Hopefully this article inspires you to build your own automated content machine.
And if you need help deciding which Skills to equip your agent with, I build a Skill Builder GPT that interviews you and gives you a custom roadmap.
Get it here: