Scheduled AI Blog Draft Generator and Ghost CMS Publisher in n8n

Schedule trigger → OpenAI generates post with metadata → Code extracts and formats → Ghost Admin API publishes

Who

Blog operators and Ghost site owners who want hands-off content publishing

You get

Automatically triggers every 12 hours, uses gpt-4.1-mini to generate a post with title, tags, and SEO meta, then publishes directly to Ghost via the Admin API

You need

n8n · schedule · openai · http-request · code

Steps

1. Set up the Schedule Trigger

   Add a Schedule Trigger node. Set the rule to run every 12 hours (field: hours, hoursInterval: 12). You may set triggerAtMinute (e.g., 34) to control the exact minute. Save the node as the workflow entry point.

2. Configure the OpenAI generation node

   Add an OpenAI node (@n8n/n8n-nodes-langchain.openAi) and select your openAiApi credential. Choose model gpt-4.1-mini-2025-04-14 (or switch to gpt-4.1 / gpt-4). Enter the prompt in Messages to request the blog body, title, SEO meta description, and 5 tags.

   Execution prompt · Source text。OpenAI node messages.values[0].content field

   Copy prompt

   Write a high-quality blog post on a creative or thought-provoking topic. The tone should be engaging and immersive. Length: 2–4 paragraphs.
   
   Then add a brief paragraph offering an alternative perspective or logical counterpoint.
   
   Finally, generate:
   - Blog post title
   - Meta description
   - 5 tags

3. Extract content and metadata (Code node)

   Add a Code node connected after OpenAI. Use JavaScript to parse the model output with regex, extracting values after **Blog Post Title:**, **Meta Description:**, and **Tags:**; separate body from metadata; wrap the body in a centered HTML div preserving line breaks; return a JSON object with title, html, status (published), tags array, and meta_description.

   Execution prompt · Source text。First Code node (id=4c460c9f-0074-4b9a-be5c-f0cfb8bae9bd) jsCode field

   Copy prompt

   const content = $json.message.content;
   
   // Extract Blog Post Title
   const titleMatch = content.match(/\*\*Blog Post Title:\*\*\s*(.+)/i);
   const title = titleMatch ? titleMatch[1].trim() : 'Untitled';
   
   // Extract Meta Description
   const metaMatch = content.match(/\*\*Meta Description:\*\*\s*(.+)/i);
   const meta = metaMatch ? metaMatch[1].trim() : '';
   
   // Extract Tags
   const tagsMatch = content.match(/\*\*Tags:\*\*\s*(.+)/i);
   const rawTags = tagsMatch ? tagsMatch[1] : '';
   const tags = rawTags.split(',').map(t => t.trim()).filter(t => t);
   
   // Strip metadata from the body
   const body = content
     .split('\n\n')
     .filter(p => !/\*\*Blog Post Title:\*\*/i.test(p) && !/\*\*Meta Description:\*\*/i.test(p) && !/\*\*Tags:\*\*/i.test(p))
     .join('\n\n');
   
   // Wrap in clean HTML
   const html = `<div style="text-align:center; white-space:pre-line;">${body.trim()}</div>`;
   
   // Return Ghost-compatible JSON
   return [
     {
       json: {
         title,
         html,
         status: 'published',
         tags,
         meta_description: meta
       }
     }
   ];

4. Format as Ghost mobiledoc (Code1 node)

   Add a second Code node receiving the previous JSON. Embed the HTML into a mobiledoc structure (version: 0.3.1, cards: [['html', {html: htmlContent}]], sections: [[10, 0]]); convert the tags string array into {name: tag} objects; assemble a posts array with title, mobiledoc (stringified JSON), status, and tags; pass output to the HTTP Request node.

   Execution prompt · Source text。Code1 node (id=cb384074-daf5-4945-adf6-40657498493d) jsCode field

   Copy prompt

   const title = $json.title;
   const status = $json.status || 'published';
   const tags = ($json.tags || []).map(tag => ({ name: tag }));
   
   // Already wrapped with styling and line breaks — use as-is
   const htmlContent = $json.html;
   
   const mobiledoc = JSON.stringify({
     version: '0.3.1',
     atoms: [],
     cards: [['html', { html: htmlContent }]],
     markups: [],
     sections: [[10, 0]]
   });
   
   return [
     {
       posts: [
         {
           title,
           mobiledoc,
           status,
           tags
         }
       ]
     }
   ];

5. Publish to Ghost (HTTP Request node)

   Add an HTTP Request node with POST method. Replace the URL with your actual Ghost domain: https://yourdomain.com/ghost/api/admin/posts/. Enable sendBody, set body type to JSON, and enter {{ JSON.stringify($json) }}. Add header Content-Type: application/json. Set Authentication to predefinedCredentialType, choose ghostAdminApi, and select your configured Ghost Admin API credential.

6. Save and activate the workflow

   Verify connections: Schedule Trigger → OpenAI → Code → Code1 → HTTP Request1. Confirm the Ghost URL and credentials are replaced with real values. Save the workflow and toggle Active. For first-time testing, click Execute Workflow and check your Ghost admin for the new post. [First-test safety] Before the first test, set status to draft in the first Code node, or temporarily disconnect the HTTP Request node; restore publishing only after the output shape looks correct.

You are an n8n workflow setup assistant. The user wants to recreate the "Scheduled AI blog draft generator and Ghost CMS publisher" template. Assist in this order:
1. Confirm the user has access to an n8n instance, plus an OpenAI API Key and a Ghost Admin API Key.
2. Guide them to import workflow.json or build the nodes manually.
3. Emphasize replacing the HTTP Request URL from yourcustomurl.com to the actual Ghost domain.
4. Recommend a manual execution before scheduling to verify the post appears in Ghost.
5. If OpenAI output format deviates from the regex expectations, advise adjusting the match rules in the Code node.