AI Content Pipeline for n8n: Notion Ideas to Drafts

A content backlog full of one-line ideas is worthless until someone sits down and writes. This n8n workflow does the sitting down for you: every morning it picks one idea marked Ready in Notion, produces an outline, writes a full draft, and hands it back with the status set to Review. A human still edits and publishes — the blank-page phase is just gone.

What this workflow does

• Schedule trigger fires daily at 7:00
• Notion node queries your content database for one page with Status = Ready
• OpenAI outline step (gpt-4o-mini) returns JSON: working title, target keyword, 5-7 H2 sections with bullet points
• OpenAI draft step (gpt-4o) writes a 900-1,200 word markdown draft from that outline — short paragraphs, concrete examples, no filler
• Notion append adds the draft to the idea’s page body
• Notion update sets the status to Review

The two-step chain matters: outlining and drafting in separate calls produces noticeably better structure than one mega-prompt, and you can review or edit the outline stage independently.

Prerequisites

• n8n instance with the LangChain OpenAI node (bundled in current versions)
• OpenAI API key with access to gpt-4o-mini and gpt-4o
• A Notion integration token, shared with your content database
• A Notion database with at least a Name title property, a Status property (with Ready and Review options), and optionally a Notes rich-text property for the angle

How to import

1. Download the JSON from the box at the top.
2. In n8n: Workflows → Import from File.
3. Attach your credentials where you see YOUR_NOTION_CREDENTIAL and YOUR_OPENAI_CREDENTIAL, and set YOUR_NOTION_DATABASE_ID in the “Get Ready Ideas” node.
4. Run it once manually with a test idea marked Ready, check the draft lands on the page, then activate the schedule.

What to customize

• The audience line in both system prompts — the default targets “operations and marketing managers evaluating automation tools”; change it to your reader or the drafts will sound generic
• Word count and tone rules in the draft prompt
• Schedule — daily is aggressive for small teams; weekly batches work fine
• The limit: 1 on the Notion query — raise it to draft multiple ideas per run (each loops through both AI steps)
• Status names if your board uses “To Write” / “In Review” instead

The full editorial-workflow version of this setup, including SEO briefs and a publishing step, is covered in the AI content pipeline tutorial.

Cost per run

This is the most token-hungry template in our library because of the gpt-4o draft step: expect roughly 2,500-4,000 output tokens per draft plus the outline call. As of mid-2026 that lands around 3-6 cents per article — still absurdly cheap against a freelancer’s first-draft rate. Swap the draft model to gpt-4o-mini to cut the cost by an order of magnitude at some quality loss.