Build an AI Content Pipeline with n8n: From RSS to Blog Post

One of our most-used internal workflows at Vorlux AI is the Content Pipeline: an n8n workflow that monitors RSS feeds, summarises articles with a local LLM, generates blog post drafts, and queues them for review. Here’s how to build your own.

What You’ll Build

A workflow that:

1. Monitors 3-5 RSS feeds for AI industry news
2. Filters for relevance using keyword matching
3. Summarizes each article using Ollama (local LLM)
4. Generates a blog post draft with SEO metadata
5. Saves to a Google Sheet or Notion database for review

Estimated API cost: EUR 0: everything runs locally.

What You’ll Build: The End Result

Before diving into the setup, here is exactly what the finished pipeline produces every day, without any manual effort:

• Automated monitoring of 3-5 RSS feeds, polling every 30 minutes for fresh AI industry news
• Smart filtering that discards irrelevant articles using keyword matching, so only on-topic content enters your pipeline
• AI-generated summaries of each article in 3 concise bullet points, written for a business audience
• Complete blog post drafts of 300+ words with practical takeaways, ready for human review and light editing
• SEO metadata including suggested title, tags, and meta description for each draft
• Organized review queue in Google Sheets or Notion, with status tracking (Draft / Reviewed / Published)
• Full audit trail showing the original source URL for every generated piece, ensuring proper attribution

In practice, this pipeline generates 5-15 draft blog posts per week depending on how many feeds you monitor and how broad your keywords are. A single human reviewer can process the entire weekly batch in under 30 minutes, turning raw AI output into publish-ready content. That replaces approximately 8-12 hours of manual research, reading, summarizing, and writing per week.

flowchart LR
    RSS["📡 RSS Feeds"] --> FILTER["🔍 Keyword<br/>Filter"]
    FILTER --> OLLAMA["🤖 Ollama<br/>Summarize"]
    OLLAMA --> FORMAT["📝 Format<br/>Draft"]
    FORMAT --> SHEETS["📊 Google Sheets<br/>Review Queue"]
    SHEETS --> PUBLISH["🚀 Publish"]
    
    style RSS fill:#F5A623,color:#0B1628
    style OLLAMA fill:#059669,color:#FAFAFA
    style PUBLISH fill:#059669,color:#FAFAFA

Prerequisites

• n8n installed (self-hosted or cloud)
• Ollama running locally with llama3.1:8b or similar
• RSS feed URLs for your target sources

Step 1: RSS Feed Trigger

Create an RSS Feed Trigger node. Add your sources:

• https://news.ycombinator.com/rss (Hacker News)
• https://techcrunch.com/category/artificial-intelligence/feed/ (TechCrunch AI)
• https://feeds.feedburner.com/TheHackersNews (Security news)

Set the polling interval to every 30 minutes.

Step 2: Keyword Filter

Add an IF node to filter articles. Check if the title or description contains your target keywords:

{{$json.title.toLowerCase().includes('ai') || 
  $json.title.toLowerCase().includes('llm') || 
  $json.description.toLowerCase().includes('edge computing')}}

This ensures you only process relevant articles, saving compute time.

Step 3: AI Summarization with Ollama

Add an HTTP Request node pointing to your local Ollama instance:

• URL: http://localhost:11434/api/generate
• Method: POST
• Body:

{
  "model": "llama3.1:8b",
  "prompt": "Summarize this article in 3 bullet points for a business audience:\n\nTitle: {{$json.title}}\n\nContent: {{$json.description}}",
  "stream": false
}

Since Ollama runs locally, this costs EUR 0 per request with ~12ms latency.

Step 4: Blog Post Generation

Add another Ollama call to generate a full blog post draft:

{
  "model": "llama3.1:8b",
  "prompt": "Write a 300-word blog post about this topic for a Spanish SME audience interested in AI deployment. Include practical takeaways.\n\nTopic: {{$json.title}}\nSummary: {{$node['Ollama Summary'].json.response}}",
  "stream": false
}

Step 5: Save for Review

Add a Google Sheets or Notion node to save the output:

• Title
• Original URL
• AI Summary (3 bullets)
• Draft blog post
• Suggested tags
• Status: “Draft”

Download This Workflow

We’ve built a production-ready version of this pipeline that includes error handling, deduplication, and multi-language support.

Download the n8n workflow JSON: import directly into your n8n instance.

Running Costs

Compare this to using GPT-4o API for the same workflow: approximately EUR 50-200/month depending on volume.

Scaling and Quality Assurance

Once your pipeline is running, consider these enhancements to maintain content quality at scale:

• AI Evaluations: n8n’s built-in AI Evaluations feature lets you run test datasets through your workflow and measure output quality scores automatically, essential for catching quality drift before it reaches production.
• Multi-model routing: Use a Code node to route tasks to different models based on complexity. Simple summaries go to Phi-4 (fast), research-heavy content goes to Llama 3.3 70B (deep reasoning).
• Human-in-the-loop: Add an approval step via email or Slack before the publish node. Automation handles 90% of the work; a human validates the final 10%.
• Scheduled runs: Use n8n’s built-in Cron trigger instead of webhooks for daily or weekly content batches. This is more reliable than webhook-based triggers for content pipelines.

Next steps

• Set up your RSS Feed Trigger with your chosen sources
• Configure the IF node to filter for your specific keywords
• Point your HTTP Request node to your local Ollama instance
• Add a node to save your drafts to Google Sheets or Notion
• Review the workflow JSON to implement error handling and deduplication

Related reading

• Build a Self-Improving AI Agent with n8n: The Learning Loop Workflow
• Guide Workflows Gstack Workflow
• n8n Workflow Automation: Complete Guide for AI-Powered Businesses

Related Resources

• 230 Downloadable Workflows: browse our full n8n + ComfyUI library
• ROI Calculator: compare cloud vs local AI costs
• Hardware Catalog: 13 devices for running n8n + Ollama locally
• Edge AI vs Cloud Cost Analysis: detailed cost breakdown

Need help implementing AI workflows in your business? Schedule a free consultation to design a pipeline tailored to your needs.

Work with us

We size the model and the machine by measuring, not by guessing. If you want to see your own task running on real hardware, book a 15-minute call or see how we work in consulting.