YouTube Comment Analyzer

You just published a video and the comments are piling up. YouTube Comment Analyzer reads them all — fetching in bulk, processing in batches — and tells you exactly how your audience feels: what percentage loved it, what frustrated them, which comments best represent each camp, and crucially, which ones haven't been replied to yet so you know where to engage first.

Usage

Send a message with a YouTube video URL or ID:

Message	Effect
https://www.youtube.com/watch?v=VIDEO_ID	Analyse top 100 comments
https://youtu.be/VIDEO_ID	Analyse top 100 comments
https://www.youtube.com/shorts/VIDEO_ID	Analyse top 100 comments
VIDEO_ID	Bare 11-character ID
VIDEO_ID 200	Analyse up to 200 comments

Examples:

• "https://youtu.be/dQw4w9WgXcQ" — analyse the top 100 comments
• "dQw4w9WgXcQ 300" — go deeper with 300 comments

What the report includes

• Counts and percentages — positive / neutral / negative breakdown across all analysed comments
• Representative examples — up to 3 real comments shown per sentiment category
• Needs reply — up to 5 unreplied top-level comments flagged for creator follow-up
• Total analysed — exact count of comments fetched and classified

Sentiment definitions

Sentiment	Criteria
positive	Praise, excitement, appreciation, satisfaction
neutral	Questions, plain statements, mixed reactions, off-topic
negative	Criticism, frustration, disappointment, hostility

Environment variables

Variable	Description
YOUTUBE_API_KEY	YouTube Data API v3 key — configured at deploy time
OPENAI_API_KEY	Auto-injected by Astropods

Limitations

• Fetches top-level comments only; reply content is not analysed, only whether a reply exists.
• Comments are sorted by relevance (YouTube default); chronological ordering is not supported.
• Comment text is truncated at 300 characters per comment before sending to OpenAI.
• Videos with disabled comments will return no results.
• YouTube Data API v3 quota applies — high comment limits may consume significant quota.
• The full report is returned once all comments are processed; per-batch progress is not streamed.