How I Turned 33,567 Scraped FB Posts Into a 4.82 ROAS Ad Angle Engine
In a Post-AI World, Better Data In = Better Results Out
Most marketers still guess what angles will work.
I scraped 33,567 viral FB posts so I didn’t have to.
Using Apify, I pulled posts from niche-specific Facebook Groups tied to the product I was marketing.
After filtering, I kept the top 100 most engaged posts.
Fed them into ChatGPT.
Extracted emotional hooks, CTAs, angles.
Tested 7 of them.
Best performer? 4.82 ROAS - on day 1.
No guesswork. Just market-native insight turned into ads.
The System: How I Extracted 7 Ad Angles From 33,567 Posts
4 step process: Scrape → Sort → Analyze → Adapt
Step 1: Scrape
I used Apify’s Facebook Groups Scraper to scrape public group posts from 8 niche communities relevant to my product.
Once the scraper was set up, I ran a batch crawl collecting 33,567 posts across 9 different Facebook groups with timestamps, likes, comments, and post content.
Here’s what it looked like inside Apify:
You can export this in JSON, CSV, or straight to Google Sheets, ready for cleaning and sorting.
Step 2: Sort
ChatGPT can’t process 33,567 posts at once so I needed to trimmed the input.
I sorted the data by total engagement (likes + comments + shares), then filtered down to the top 100 posts, each with at least 500 likes.
This gave me a high-signal shortlist. Real posts, with real traction, written in the language of the niche.
Step 3: Analyze
With the top 100 posts selected, I fed each one — text + image — into ChatGPT.
For every post, I asked it to extract:
The emotional angle (e.g. fear, hope, nostalgia)
The hook structure
The CTA style (explicit or implied)
Including the visual helped ChatGPT understand tone, context, and why the post may have performed.
This turned raw community content into structured, testable ad angles.

Step 4: Adapt
Once ChatGPT broke down the angles, I asked:
“How can I turn this post into a high-converting ad for [my product]?”
ChatGPT then rewrote each post while keeping the emotional core, but adapting the copy to fit my offer, product, and funnel.
I repeated this across all 100 inputs and shortlisted 7 ad angles worth testing.
The result? 4.82 ROAS on day 1.
🧠 Why I Think This Worked
Most winning Facebook ads don’t feel like ads.
They look like content you’d naturally see in your feed. Native, emotional, scroll-stopping.
By scraping niche-specific Facebook groups, I wasn’t guessing.
I was pulling real content that had already gone viral while written in the voice of the community, with visuals that blended naturally into the feed.
That gave me two key advantages:
Visual cues that didn’t look like stock ads
Ad angles based on real conversations and emotional triggers
Instead of writing from scratch, I adapted what was already working organically and made it fit the product.
⚙️ Building It Into a Repeatable System
This isn’t a one-time workflow.
It’s something I can automate to run monthly.
Use n8n to trigger a monthly job
Call the Apify API to scrape new FB group posts
Sort by engagement → keep the top 100
Feed them into ChatGPT to extract fresh ad angles
Output to Google Sheets or Airtable for review and testing
Each run gives me a new batch of angles based on what’s gone viral in the past 30 days.
No more guesswork.
Just a working system that mines your niche for proven angles, again and again.
Wrap Up: AI Isn’t the Shortcut. Data Is.
AI can’t think for you. But it can distill huge datasets into usable, testable insights if you feed it the right inputs.
That’s why data collection is now a core marketing skill.
Not just for analysts, but for every marketer using AI to write, test, or optimize.
The marketers who win won't be the ones prompting harder.
They’ll be the ones who own their inputs and build systems around them.
I test AI workflows on real marketing problems.
And turn the useful ones into 10x Playbooks.
So business owners and solo marketers doing it all themselves can skip the guesswork… and copy what actually works.
→ I drop one of these 10xPlaybooks 🚀 every week. Don’t miss the next.
John
Really impressed by this approach. Super practical breakdown of how AI and automation can actually be applied to solve a specific business challenge
💥💥 Amazing Idea 💥💥