What This Automation Is About (Big Picture)
This is Version 2 of an end-to-end SEO blog automation built in n8n.
The goal is simple:
Give one topic + keyword → get a fully researched, human-sounding, SEO-ready blog post automatically published everywhere.
What makes this version powerful is that it goes far beyond basic AI writing:
- Deep SERP + widget research
- Structured outlines (no random AI articles)
- Word-limit enforcement
- Multi-layer humanization
- AI detection + re-humanization
- High-quality image generation
- Automatic publishing to Google Drive, Notion, and WordPress
What This Automation Does
This is an end-to-end SEO blog automation built in n8n.
You give it:
- A topic
- A keyword
It automatically creates a fully researched, human-sounding, SEO-ready blog and publishes it to Google Drive, Notion, and WordPress.
Step 1: Trigger the Automation (Schedule Node)
The workflow starts with a Schedule Trigger.
Example use:
- Run every day at 7:00 AM
- Or once per week
- Or manually for testing
This allows the system to generate content while you sleep.
Step 2: Read Pending Topics from Google Sheets
Next, the workflow:
- Fetches all rows from Google Sheets
- Filters only rows with status = pending
- Loops over each topic one by one
This prevents duplicate work and keeps everything organized.
Step 3: Set Core Variables (Set Node)
For each topic, the system stores:
- Title
- Keyword
These become the single source of truth for the entire workflow.
Everything downstream references these values.
Step 4: Keyword & Intent Research (AI Agent + Tools)
This is where the system gets smart.
What happens here:
An AI Agent is used for research, not writing.
It analyzes:
- Search intent
- Writing style
- Writing tone
- Target audience
- Goal of the article
- Semantic keyword insights
- Hidden ranking opportunities
Tools connected to the agent:
- Google Sub API (Pulls SERP widgets like videos, FAQs, discussions, summaries)
- Wikipedia tool (Adds factual grounding and context)
This ensures the article matches how Google expects the topic to be written, not just what sounds good.
Step 5: Deep SERP Research (Perplexity API)
Next, the workflow runs Perplexity (Sonar Pro model).
Configuration highlights:
- Searches up to 10 pages
- Pulls:
- Top ranking articles
- URLs
- Snippets
- Competitive angles
Why this matters:
- AI now writes with real competitor awareness
- No hallucinated facts
- Strong topical authority
Step 6: Clean & Merge Research
Perplexity + Sub research outputs are:
- Cleaned
- Normalized
- Merged into one structured research block
This becomes the knowledge base for writing.
Step 7: Generate a Refined SEO Title
A new AI Agent (Claude):
- Refines the blog title
- Explains why the title is optimized
- Ensures high CTR and SEO friendliness
Output is stored using a structured output schema so nothing breaks downstream.
Step 8: Define Key Takeaways
Another AI Agent generates:
- Clear reader takeaways
- What the reader should learn or do after reading
This keeps the article focused and useful.
Step 9: Create a Structured Blog Outline
The system now:
- Builds a proper outline (H1, H2, H3)
- Prevents random or repetitive content
- Ensures logical flow
This step is critical for long-form SEO quality.
Step 10: Write the First Draft
Using the outline + research:
- The AI writes the full article
- Example length: 2000 words (customizable)
This is the raw draft, not final yet.
Step 11: Word Count & Character Check (Code Node)
A Code Node calculates:
- Word count
- Character count
- Limit exceeded or not
Why this matters:
- Keeps articles consistent
- Prevents bloated AI content
Step 12: Auto-Shorten if Needed (AI Agent Loop)
If the article exceeds the limit:
- A custom AI agent rewrites it
- Brings it under the target word count
- The system checks again
- Loops until valid
This guarantees compliance without manual editing.
Step 13: Humanization Layer (Custom AI Agent)
Now the system removes “AI-sounding” patterns.
The humanizer:
- Removes clichés
- Removes buzzwords
- Avoids metaphors
- Avoids over-descriptive fluff
- Removes typical ChatGPT phrases
Goal:
Make the article sound like a
real human expert
Step 14: AI Detection Check (External API)
A third-party AI detection API checks:
- Is this content detectable as AI?
If YES:
- Run humanization again
If NO:
- Move forward
This adds an extra safety layer for SEO.
Step 15: Convert to Clean HTML
Once finalized:
- The article is converted into clean HTML
- No commentary
- No markdown
- SEO-friendly structure
A code node formats:
- Headings
- Paragraphs
- Lists
- Styling
Step 16: Generate SEO Metadata
An AI agent generates:
- SEO title
- Meta description
- Image prompt for the article
This uses a cheap model because it’s a logic task, not creative writing.
Step 17: Image Generation (Nano Banana via OpenRouter)
The image pipeline:
- Uses Nano Banana (Google image model)
- Forced 3:1 aspect ratio
- Reference image included
- Prompt is generated automatically
Image is:
- Converted to Base64
- Uploaded to ImgBB
- URL stored for publishing
Step 18: Prepare Final Post Object
Everything is now bundled together:
- Final title
- SEO title
- SEO description
- HTML body
- Featured image URL
- Original keyword
This becomes the final “publishable post”.
Step 19: Omni-Channel Publishing
The article is automatically published to:
Google Drive
- Create document
- Update with HTML body
- Make public
- Save URL
Notion
- Convert HTML → Notion blocks
- Validate each block
- Insert page content
- Add metadata fields
WordPress
- Upload image
- Create post
- Assign featured image
- Publish
(Each platform can be enabled or disabled.)
Step 20: Update Google Sheet & Notify
Finally:
- Status is updated to Completed
- Article links are saved
- SEO metadata is written back
- Optional email notification sent
The loop closes cleanly.
Final Outcome
From one row in Google Sheets, the system delivers:
- A researched SEO article
- Human-sounding content
- Optimized metadata
- Custom images
- Published everywhere
- Zero manual work
This is no longer “AI writing”.
This is a full-scale content production system.