Back to Templates
How it works
- A form trigger accepts any YouTube URL: youtube.com/watch?v=ID, youtu.be/ID, youtube.com/embed/ID, or just a plain video ID
- A code node normalises the URL and extracts a clean video ID for the YouTube Data API v3
- Calls the API to fetch video stats (title, description, tags, views, likes, comment count)
- Fetches comments from the videos commentThreads endpoint, capped by the form's Max Comments to Process field
- Flattens each comment into its own item, then merges two ranked lists deduped by commentId: the 5 newest comments by publishedAt and the 5 highest-engagement comments by likes plus replies
- A GPT-4 agent writes a natural, human-sounding reply for each top comment, responding as a fellow viewer (not as the channel owner)
- In parallel, a second GPT-4 agent writes one original top-level comment based on the videos title and description
- A structured output parser guarantees clean JSON for both agents
- Replies are merged with the top-level comment by videoId, normalised into row shape, and appended to Google Sheets - one row per reply
Set up steps
- Setup takes about 5 to 10 minutes
- Connect a YouTube Data API OAuth2 credential (used by both the Get Video Statistics and Get Video Comments nodes)
- Connect an OpenAI API credential (defaults to gpt-4.1-mini; swap to any model you prefer)
- Connect a Google Sheets OAuth2 credential and pick your destination spreadsheet plus tab in the Save to Google Sheets node
- Create a tab with these column headers in row 1: videoId, videoUrl, videoTitle, commentAuthor, commentText, myReply, myVideoComment, selectionType, videoTags, viewCount, engagementScore, videoDescription, likeCount, commentCount
- Optional: tweak the system prompts inside each AI agent to match your niche, voice, and brand
- Submit a YouTube URL via the form trigger to test
- Detailed per-step notes live inside the workflow as sticky notes