PlainAgent / Templates / YouTube to blog post agent
Content templateYouTube to blog post agent
A YouTube to blog post agent takes one of your videos and writes a first blog draft from it: a title, a meta description, the post, and a list of every claim, number and name to check, each with the minute it comes up. You edit and publish; the agent never posts on its own.
Turn one video you made into a blog draft in your own points, plus a fact-check list with timestamps.
You paste a video link (or its transcript) into a form or a chat.
Always. Every draft goes to you as a document to check and edit; nothing is published automatically.
- The video itself (Gemini can take a public YouTube link) or its transcript (any other model)
- Optional notes: the audience, a keyword, the length you want
- Read the video or the transcript
- Create one draft document (Google Docs, Notion or your CMS as a draft)
- Only use what is in the video and your notes: no added facts, prices or quotes.
- The speaker’s points in new words, not a word-for-word transcript.
- Only run videos you made or have the rights to repurpose.
How to build it
- Decide how the model gets the words. Gemini can watch a public YouTube video from its link. Claude and ChatGPT can’t take a YouTube link as video input, so give them the transcript; for your own videos, YouTube Studio lets you download the captions.
- Make a form or a chat entry point with two fields: the link (or transcript) and optional notes.
- Paste the system prompt below into the AI step and ask for JSON, so the title, description, draft and check list arrive as separate fields.
- Give the AI step room to write: a 600 to 1,000 word draft needs an output limit of a few thousand tokens. A short limit cuts the draft off mid-sentence.
- Write the result into a new document as a draft, with the check list at the bottom.
- Check every line of the check list against the video before you publish, then add headings and links in your editor.
System prompt
Replace the parts in {BRACES}. Works in most builders (Custom GPTs, Claude Projects, Copilot Studio, n8n/Zapier/Make AI steps).
You are an editor turning a video into a first blog draft for the video's own creator. Use the {VIDEO OR TRANSCRIPT} below.
Notes from the creator (may be empty): {NOTES}
Return JSON only, with these keys:
title: a blog title under 60 characters that says what the reader gets
meta_description: 120 to 155 characters
draft: the post in plain text, 600 to 1,000 words. Start with a 2-sentence answer to the video's main question. Put each section heading on its own line. Use the speaker's own points and examples, in your own words: do not copy long passages word for word.
check_these: a list of objects with timestamp (MM:SS) and claim, for every number, price, date, name and factual claim in the draft.
Use only what is in the video and the notes. If something is unclear (a name you couldn't hear, a number that went by fast), put it in check_these instead of guessing.
Test before you switch it on
- Run it on a video you know well, and check every number and name in the draft against the check list and the video.
- Run it on a video with a price or a date shown only on screen: it should appear in the check list, with the right minute.
- Run it on your longest video and check the draft isn’t cut off and the JSON is complete.
- Search the draft for a sentence you said word for word. Long copied passages mean the prompt isn’t being followed.
- Try an unlisted or private video: the run should fail with a clear message, not write an empty draft.
What 40 public YouTube-to-blog templates show
In our study of 13,574 public n8n templates (snapshot 2026-10-20), 40 templates have “youtube” plus “blog”, “article” or “post” in their name. Some of them turn a video into social posts rather than a blog, so read this as a rough picture:
| The template… | Templates | Share |
|---|---|---|
| Uses an AI model | 36 | 90.0% |
| Makes an HTTP Request call | 33 | 82.5% |
| Uses OpenAI | 21 | 52.5% |
| Has a YouTube node | 11 | 27.5% |
| Uses Google Gemini | 6 | 15.0% |
| Uses Anthropic’s Claude | 5 | 12.5% |
| Has a WordPress node | 5 | 12.5% |
| Writes to Google Docs | 4 | 10.0% |
| Has an Error Trigger | 0 | 0.0% |
Most builders get the words first: 8 in 10 call an outside service over HTTP, usually a transcript or download tool, then hand the text to a model, most often OpenAI. Only 6 of 40 use Gemini, even though it is the one model family here that can take the YouTube link itself. Only 5 have a WordPress node, and none has an Error Trigger, so a failed run would just produce no draft without telling anyone. Keep a person between the draft and the publish button.
Four Gemini video rules that decide how you build it
- Public videos only, and free for now. Google’s video understanding docs (read 2026-11-01) say the YouTube link feature is in preview, available at no charge, and only works with public videos, not private or unlisted ones. The free tier allows up to 8 hours of YouTube video a day.
- Long videos need a big context window. The same docs say a model with a 1M-token window can take about 3 hours of video at the default low media resolution, or 1 hour at high resolution, at roughly 100 tokens per second of video at the default.
- If it starts being billed, it is cheap. Gemini 3.5 Flash-Lite costs $0.30 per million input tokens and $2.50 per million output tokens on the paid tier (Gemini API pricing, read 2026-11-01). Our arithmetic: a 15-minute video is about 90,000 input tokens plus about 2,000 output tokens, about 3 cents a video. On the free tier Google says it may use your prompts to improve its products.
- n8n’s Gemini node can’t take a YouTube link. In the n8n source we read (@n8n/n8n-nodes-langchain 2.45.6), the Gemini node’s Video → Analyze operation downloads the link and re-uploads what it got, so a YouTube link sends the web page, not the video. Call the Gemini API with an HTTP Request node instead, as our YouTube to blog draft recipe does.
Totals from the public n8n template library (snapshot 2026-10-20), computed with the same definitions as the study. Group: templates whose name has “youtube” and “blog”, “article” or “post”. A model provider counts if the template has any node from that vendor. A template can contain several nodes, so shares overlap. No individual templates are published.
New to this? Read how to build an AI agent, or generate your own spec with the agent planner. To turn the same video into short social posts instead, see the content repurposing agent.