# n8n: RSS to Social Posts with AI Summaries

> Automatically monitor RSS feeds, summarize articles with an LLM, and publish branded posts to LinkedIn, X, and a Notion archive.

**Platform:** n8n  
**Category:** n8n  
**Difficulty:** Beginner  
**Tags:** n8n, content, social, openai  
**Updated:** 2025-06-14  
**Source:** https://promptfoo.tech/workflows/n8n-rss-to-social-with-ai-summaries

## Overview
This n8n workflow watches a set of RSS feeds you care about, uses an LLM to write a concise 2–3 sentence summary with an editorial hook, and publishes the result to LinkedIn and X while archiving every item to a Notion database. It replaces the manual 'read, summarize, post' loop most operators do every morning.

## How it works
1. A schedule trigger runs every 30 minutes and fans out to multiple RSS Read nodes.
2. A Function node de-duplicates entries against a Notion database using the article URL as a key.
3. New entries are sent to an OpenAI node with a prompt that enforces tone, length, and a call-to-action.
4. The generated post is routed through Set nodes for platform-specific formatting (hashtags, line breaks, UTM tags).
5. Publish nodes push to LinkedIn and X; a final Notion node archives the article, summary, and post URL.

## Benefits
- Cuts a 45-minute daily content ritual to under 5 minutes of review.
- Consistent voice across channels because the prompt is versioned in one place.
- Full audit trail of what was published, when, and from which source.

## Use cases
- Solo founders who curate an industry newsletter feed.
- DevRel teams tracking releases across dozens of open-source projects.
- Agencies running always-on thought-leadership for multiple clients.

## Step-by-step
### Step 1: Import the workflow JSON
Copy the workflow JSON, open n8n, and use Import from File. All credentials remain empty until you wire them up.

### Step 2: Connect your feeds
Duplicate the RSS Read node for each source. Group related feeds so the downstream prompt can adapt tone per topic.

### Step 3: Set up your Notion archive
Create a database with URL, Title, Summary, Source, and Published At columns. This is your de-dup key and your public archive.

### Step 4: Tune the LLM prompt
Start with the provided system prompt. Adjust voice, forbidden words, and hashtag policy. Keep temperature at 0.4 for consistency.

### Step 5: Connect LinkedIn and X
Use OAuth via n8n's built-in credentials. Test with a private post first before enabling the schedule trigger.

### Step 6: Activate and monitor
Turn the workflow on. Check the executions tab daily for the first week and refine the prompt from real output.

## Example
```
Input article: 'OpenAI Announces GPT-5 Preview'
Generated LinkedIn post: 'OpenAI just previewed GPT-5 with a step-change in reasoning benchmarks. The interesting part isn't the score — it's the pricing signal for what's coming to production APIs in Q1. Full breakdown: <link>'
```

## FAQs
**Do I need self-hosted n8n?**

No. The workflow runs on n8n Cloud, self-hosted Docker, or the desktop app. Only difference is where credentials are stored.

**Which LLM should I use?**

GPT-4o-mini or Claude 3.5 Haiku are both cheap and strong enough for summarization. Reserve larger models for editorial rewrites, not first drafts.

**How do I avoid duplicate posts?**

The Notion de-dup step is essential. Never rely on the RSS pub date alone — many feeds re-publish items when tags change.

---
Licensed under CC-BY-4.0. Attribution: PromptFoo.tech