Long-Form YouTube with AI.

YouTube doesn't promote videos. It matches viewers with content they'll watch.

YouTube's recommendation engine has one job: keep people on the platform. It does this by tracking two key metrics for every video: click-through rate (CTR) — the percentage of people who click when they see your thumbnail — and average view duration (AVD) — how long they actually watch. A video with a 10% CTR and 50% AVD will outperform a video with a 15% CTR and 20% AVD every time, because the second video is clickbait: people click but don't stay.

This means your job is two-fold: create packaging (title + thumbnail) that earns clicks, and create content that delivers on the promise so people watch. AI helps with both. But the order matters: never create a video and then figure out the title. Create the title and thumbnail FIRST, then make the video that delivers on that promise. This is how every major YouTuber works.

AI turns 10 hours of research into 30 minutes.

Before you shoot anything, validate the idea. Prompt: "I want to make a YouTube video about [topic]. Research the top 10 performing videos on this topic from the last 12 months. What angles have been covered? What angles are missing? What titles and thumbnails got the most views? What comment themes suggest viewers want more on specific subtopics?"

For scripting, use the HIVE structure: Hook (first 30 seconds — tease the payoff, create stakes), Intro (establish credibility and preview what they'll learn), Value sections (3-5 main points, each with a mini-hook, explanation, example, and transition), End (deliver the promised payoff, CTA for likes/subscribes, tease next video).

AI scripting prompt: "Write a 10-minute YouTube script using HIVE structure on [topic]. Each value section should have a re-hook to prevent drop-off. Include [B-ROLL] and [GRAPHIC] cues for editing. Write in a conversational tone — this will be spoken, not read. Include one 'open loop' in the first minute that doesn't get resolved until the end."