Build a Viral-to-Value Content Engine with Claude and n8n

Use the n8n Apify node to trigger the workflow whenever a new trending post or video is detected. Configure the Apify 'X (Twitter) Scraper' or 'YouTube Transcript Scraper' to watch for specific keywords or high-engagement accounts.

Set the scraping frequency to once every 6 hours to balance fresh content with API costs. Map the output fields: text (for X) or transcript (for YouTube), url, and engagement_metrics.

Watch out: Ensure you set a 'Max items' limit in Apify (e.g., 5 per run) to avoid runaway costs if a keyword suddenly explodes in popularity.

Add a Function node to strip metadata, URLs, and emojis from the raw text. For YouTube, use a regex to remove timestamps and speaker names from the transcript. This ensures the LLM receives only the high-signal content.

const raw = $input.item.json.text || $input.item.json.transcript;
const clean = raw.replace(/[\u2014\u2013]/g, '-')
                 .replace(/(?:https?|ftp):\/\/[\n\S]+/g, '')
                 .trim();
return [{ json: { cleanContent: clean, sourceUrl: $input.item.json.url } }];

Watch out: If the source is a long YouTube video, the transcript might exceed Claude's context window. Truncate the text to the first 50,000 characters if necessary.

Send the cleaned content to claude-3-5-sonnet with a prompt designed to identify the 'Why It Matters' (WIM). This isn't just a summary; it's an extraction of the underlying principle that makes the content valuable to your specific audience.

Watch out: Avoid using older models like GPT-3.5 or Claude 2 for this step, as they often miss the nuance of 'thought leadership' and produce generic summaries.

Use a second HTTP Request node to call the Claude API again, this time to draft the actual LinkedIn post. Feed in the extracted thesis and 'Why It Matters' from the previous step to ensure the post is grounded in depth.

Watch out: LinkedIn's algorithm penalizes posts that look too 'AI-generated' (excessive emojis, perfect listicles). Instruct Claude to use a 'punchy, human, and direct' tone with varying sentence lengths.

Connect another Claude node to expand the viral trend into a 1,200+ word blog post. Use the 'points' extracted in Step 3 as the subheadings (H2s). This ensures the blog has the depth required for SEO and authority.

Watch out: Long-form generation can sometimes hit output token limits. If the blog feels cut off, split the generation into two steps: one for the outline and intro, and another for the body sections.