🧬 I Cloned Outrank.so in 4 Hours — Here’s How (and Why)

I love taking a great product and asking, “Could I build this myself?”
This time, the target was Outrank.so — an AI-driven SEO content engine.

Yes, I know it’s way easier (and smarter) to just support Tibo and his product.
Honestly, Outrank probably gives me a better long-term solution.

But this wasn’t about replacing Outrank.
It was about learning, experimenting, and having fun.
So I gave myself a challenge:
4 hours, one sitting, using Gemini, Google Colab, and GitHub Actions.

Here’s what happened:

✅ Step 1: First Hurdle — Just Write a Blog Post

The first script?
It spat out plain text. Useless for my site.

I needed Markdown with YAML frontmatter (title, excerpt, SEO tags).

Solution?
Gemini’s Structured Output feature:

prompt_metadata = f""" For a blog post titled: "{blog_topic}", return JSON: {"excerpt": "...", "category": "...", "tags": [...], "metaTitle": "...", "metaDescription": "...", "keywords": [...]} """

Suddenly, everything clicked. JSON → YAML → Perfect Markdown.

✅ Step 2: Give It a Brain 🧠

One blog post = cool.
A content engine that never runs out of ideas = cooler.

I added Google Sheets as the control center:
✔ Keyword Research
✔ Blog Post Ideas

When the queue runs low, Gemini auto-generates:

• 15 new keywords

• SEO-optimized blog titles

Prompt for keywords:

You are an SEO strategist. Generate 15 long-tail keywords for a SaaS offering fixed-price MVP builds in 14 days. Focus on 2025 trends & high commercial intent.

✅ Step 3: Add Brand Voice

Early drafts? Too generic.

So I baked in my Brand Brief:

• Mission: Save founders from endless dev cycles

• Core Offer: Fixed-price MVP in 14 days

• Philosophy: Speed beats perfection

Every section prompt became:

Write a 200-word section that aligns with: - Speed is an unfair advantage - Certainty > speculation Do NOT mention company name or pricing. Output in Markdown.

✅ Now, every post feels like me.

✅ Step 4: Automate Everything ⚡

Manual runs = boring.
I wanted CI/CD for blogging:

✔ GitHub Actions runs twice a day
✔ Generates post + image
✔ Pushes to repo
✔ Opens a Pull Request

Problems?

• Rate limits → Fixed by using gemini-2.5-flash for metadata, Pro for content

• Google auth in CI → Service Account + GitHub Secrets

• Git errors → Configured user & paths

✅ The Engine Is Live 🚀

Now, the system:
✔ Runs on autopilot
✔ Writes posts
✔ Generates images
✔ PR ready for merge

What started as a 4-hour experiment now saves me hours every week and scales content like crazy.

What This Means for Indie Hackers

• Automation isn’t just for dev ops. Content = leverage.

• AI can handle the boring parts if you give it structure.

• One afternoon of setup = recurring growth.

Would you build something like this for your product?
Or should I share the full Colab → GitHub Actions script?

👉 Check it live: MVPExpert.com
💬 Drop your thoughts — Would love feedback from fellow hackers.