PDF to Ghost Blog Post: n8n Automation Workflow

Upload PDF → AI extracts and structures a blog post → Publish to Ghost CMS

Who

Content teams, digital publishers, anyone needing to turn PDFs into blog posts quickly

You get

PDFs automatically converted into structured, HTML-formatted blog drafts published to Ghost

You need

n8n · ghost

Steps

1. Configure credentials

   Add two credentials in n8n: 1) OpenAI API Key (type openAiApi) for the AI node; 2) Ghost Admin API credentials (type ghostAdminApi) for publishing. Ensure the Ghost credential has permission to create posts.

2. Upload PDF and extract text

   The workflow starts with an n8n Form Trigger at path /pdf, titled 'PDF2Blog', accepting a single .pdf file. After upload, the Extract from File node extracts text with operation=pdf, binaryPropertyName pointing to the uploaded file field. You should see extracted plain text passed to the next node.

3. Configure AI generation node

   Connect gpt-4o-mini (model ID gpt-4o-mini-2024-07-18, responseFormat set to json_object) as the language model, and Structured Output - JSON as the output parser with schema example {"title":"title","content":"content"}. The Create Structured Blog Post node uses conversationalAgent, promptType=define, and binds the above model and parser. retryOnFail is enabled.

   Execution prompt · Source text。Create Structured Blog Post node systemMessage field (options.systemMessage)

   Copy prompt

   =Analyze the provided PDF article and create a compelling blog post that will be returned in JSON format with two fields: "title" and "content". Follow these specifications:
   
   ## Title Requirements
   - Create an engaging, SEO-friendly title under 10 words
   - Must not contain a colon
   - Should capture the article's essence
   - Will be formatted as an H1 in the content
   
   ## Content Structure
   - Introduction (150-200 words)
     * Compelling hook
     * Topic context and importance
     * Preview of main points
   
   - Main Content (6-8 chapters)
     * Each chapter requires:
       - Relevant H2 heading
       - 300-400 words of unique content
       - Specific topic focus
       - Source material quotes/data
       - Smooth transitions
   
   - Conclusion (200-250 words)
     * Key takeaways
     * Final thoughts/implications
   
   ## Formatting Guidelines
   - Use proper HTML tags throughout
   - Structure with <p> tags for paragraphs
   - Include appropriate spacing
   - Use <blockquote> for direct quotes
   - Maintain consistent formatting
   - Write in clear, professional tone
   - Break up long paragraphs
   - Use engaging subheadings
   - Include transitional phrases
   
   The content should be original, avoid direct copying, and maintain a consistent voice throughout. The final JSON response should contain only the title and content fields, with the content including all HTML formatting.

4. Parse output and validate

   The Separate Title & Content code node receives AI output, validates the input[0].json.output.output structure, extracts title and content, and strips any existing <h1> tag from content. If title or content is missing, it returns a failure object with error/message/timestamp. Then the If node checks both title and content are non-empty: if true, flow proceeds to Post to Ghost; otherwise to Do Nothing.

   Execution prompt · Source text。Separate Title & Content node jsCode field

   Copy prompt

   try {
     // Check if input exists and has the expected structure
     const input = $input.all();
     if (!input || !input.length) {
       throw new Error('No input data received');
     }
   
     const firstItem = input[0];
     if (!firstItem || !firstItem.json || !firstItem.json.output || !firstItem.json.output.output) {
       throw new Error('Invalid input structure: missing required properties');
     }
   
     const output = firstItem.json.output.output;
     
     // Validate title exists
     if (!output.title) {
       throw new Error('Missing title in output');
     }
   
     // Validate content exists
     if (!output.content) {
       throw new Error('Missing content in output');
     }
   
     const title = output.title;
     const content = output.content.replace(/<h1>.*?<\/h1>/s, '').trim();
   
     // Validate final content is not empty after processing
     if (!content) {
       throw new Error('Content is empty after processing');
     }
   
     console.log('Successfully processed content');
   
     console.log(title)
     console.log(content)
     
     return { title, content };
   
   } catch (error) {
     // Log the error for debugging
     console.error('Error processing content:', error.message);
     
     // Return a graceful failure object
     return {
       error: true,
       message: error.message,
       title: '',
       content: '',
       timestamp: new Date().toISOString()
     };
   }

5. Publish to Ghost

   The Post to Ghost node uses Ghost Admin API with operation=create, mapping title and content from the previous step's output. On success, a new draft blog post should appear in the Ghost admin panel with title and content populated.

You are reproducing the n8n template 'PDF2Blog'. First confirm you have an OpenAI API Key and Ghost Admin API credentials, then import the workflow JSON into n8n or build manually per the steps above. The main prompt is attached in step 3 prompts; paste it directly into the Create Structured Blog Post node's systemMessage. Test with a short PDF first to verify JSON output structure before processing long documents.