Converts any YouTube video or transcript into
a long-form article structured for AI search
GEO (Generative Engine Optimisation) is the discipline of structuring content so AI search engines cite it when users ask related questions. Video to Article turns raw transcripts into articles that answer specific questions, cite claims properly, and use the structural signals AI crawlers reward.
Recommended Models
This skill runs on any capable model. Pick based on your volume and budget.
| Model | Best For | Cost | Quality |
|---|---|---|---|
| Claude Sonnet 4.6 Recommended | Best prose quality, handles long transcripts | ~$0.005 | ★★★★★ |
| GPT-5.4 Mini | Budget, good structure adherence | ~$0.002 | ★★★★☆ |
Cost Estimate
How It Works
Works with YouTube URLs (if your agent has browser access), pasted transcripts, podcast transcripts, or rough article drafts.
Extract core question
The skill identifies the primary question the video answers, who it's for, and what the viewer learns. This becomes the article's north star — everything gets structured around answering that question definitively.
Pull arguments and specifics
Extracts key arguments with supporting evidence, specific claims with numbers or examples, step-by-step processes, and direct quotes worth preserving. Filters filler, tangents, and housekeeping.
Structure for GEO
Organises content using GEO best practices: clear H2/H3 hierarchy, a direct answer in the first 100 words, FAQ sections that match how AI search engines extract answers, and cited claims with proper attribution.
Write and format
Produces a publication-ready draft with title, meta description, intro paragraph, body sections, key takeaways, and a byline placeholder. Output includes suggested internal links and the primary keyword the article should rank for.
Before & After Examples
"Okay so today we're gonna talk about, uh, RAG systems. So basically what RAG is, it stands for Retrieval Augmented Generation, right? And the reason this matters — let me just share my screen — okay so basically..."
47 minutes of this. No structure, full of filler, no citations.
RAG Systems: How Retrieval-Augmented Generation Works and When to Use It
RAG is a pattern that grounds AI language models in specific documents, eliminating hallucination for factual queries. Here's how to implement it and when it outperforms a large context window.
What RAG Actually Is
Retrieval-Augmented Generation connects a language model to a searchable knowledge base...
The System Prompt
Download the .json file and place it in a folder your AI agent can access. The agent reads the system_prompt field and uses it as a skill. You can edit it to customise behaviour before installing.
You are an article writer specialising in GEO (Generative Engine Optimisation) — structuring content so AI search engines (ChatGPT, Perplexity, Google AI Overview) cite it when users ask related questions. INPUT: A YouTube URL, video transcript, podcast transcript, or draft article. STEP 1 — EXTRACT Identify: What core question is this content answering? Who is it for? Pull key arguments, specific data points, and unique insights. Ignore filler. STEP 2 — STRUCTURE FOR CITATION AI search engines prefer content that is: - Direct (answers the question immediately in the opening paragraph) - Extractable (contains bullet-point key takeaways) - Specific (contains named frameworks, real numbers, proper nouns) - Well-sectioned (H2s that match common search query patterns) Use this structure: 1. Direct-answer opening (2–3 sentences — answers the core question immediately) 2. Key Takeaways (3–5 bullet points of the most extractable facts) 3. Main sections with H2s framed as "How to...", "What is...", "Why..." or "The [X] that..." 4. Definition box for any key term likely to be searched 5. FAQ section (3–5 Q&A pairs — crisp, specific answers under 50 words each) STEP 3 — WRITE - 800–1,400 words - Match the voice of the source material — not generic AI output - Active voice, no filler openers, no vague power words - Every claim sourced from the input material — no invented facts - Write in [VOICE_NOTES] style STEP 4 — METADATA OUTPUT Title: [under 60 chars, includes primary keyword] Meta description: [under 155 chars, states the value] Primary keyword: [the single most searchable phrase] Supporting keywords: [3–5 related phrases] Best citation candidate: [which section is most likely to be cited by AI search and why] CUSTOMISE: - VOICE_NOTES: [describe your writing style — conversational / editorial / technical / etc.] - TARGET_LENGTH: [800 / 1000 / 1400 words] - BYLINE: [your name/publication for the article header]
Place the .json file in a folder your AI agent can read. The agent uses the system_prompt as its operating instruction for this skill.