ChatGPT Prompts for YouTube Scripts: Hook, Retain, and Convert
Scripting a 10-minute YouTube video by hand takes most creators 2–4 hours. With the right ChatGPT prompts, that drops to under an hour — and the structure is tighter than most people write on their own. The difference isn't magic; it's knowing how to give the model a role, a context, and a format instead of just a topic.
This post covers the complete scripting stack: a hook generator (with 5 variants to choose from), a retention-optimized structure, copy-paste prompts for tutorials, list videos, Shorts, and B-roll cue injection, plus a real before-and-after demo of a weak intro turned into something that actually stops the scroll. If you also want help with ChatGPT prompts for marketing, the same role-context-task-format framework transfers directly.
Why Your Script's First 15 Seconds Decide Everything
The first 15 seconds of a YouTube video determine whether a viewer stays or leaves. ChatGPT can generate 5 hook variants for any topic — a stat, a provocative question, a result-first tease, or a pattern interrupt — so you pick the strongest one rather than going with your first instinct. Viewers who get past 30 seconds are far more likely to hit the halfway mark, so the hook is the highest-leverage sentence in the entire script.
Most creators open with something like "Hey everyone, welcome back to the channel." That's 6 words spent on nothing. A hook does one of four things in the first sentence: it states a surprising result, asks a question the viewer is already asking themselves, opens in the middle of action, or makes a counterintuitive claim. ChatGPT is surprisingly good at generating these variations — the key is telling it which patterns to use.
Hook pattern comparison
| Hook type | Example (cooking channel) | Why it works |
|---|---|---|
| Result-first | "I made restaurant-quality carbonara in 12 minutes — here's exactly what I did." | Viewer knows the payoff immediately |
| Provocative question | "Why does every pasta tutorial online get this one step wrong?" | Creates an open loop the viewer has to close |
| Pattern interrupt | "Stop. Don't add cream. I'll explain in 30 seconds." | Breaks the expected format, forces attention |
| Counterintuitive claim | "The most expensive ingredient in carbonara is the one most people skip." | Challenges assumptions, earns the click |
Ask ChatGPT to generate one of each for your topic and pick whichever fits your channel voice. It takes about 20 seconds. You can also optimize your video's discoverability with an AI-optimized title — same prompt logic, different output format.
The YouTube Script Structure That Maximizes Watch Time
The standard YouTube retention structure is: Hook (0–15 sec) → Preview (15–30 sec) → Chapter body with chapter-transition mini-hooks → Recap → CTA. Give ChatGPT all five labels and specify word counts per section. This mirrors what top-performing educational and commentary channels use to sustain 40%+ average view duration across long-form videos.
The key detail most creators miss is the chapter-transition mini-hook. At the end of each body chapter, one sentence previews what's coming next — "But the part that actually surprised me is in the next section." That single line resets the viewer's reason to stay. Ask ChatGPT to include it at every chapter boundary; it does it reliably when instructed.
Script time budget: 10-minute video
| Section | Timing | Word count (at 140 wpm) | Purpose |
|---|---|---|---|
| Hook | 0:00–0:15 | 30–40 words | Stop the scroll, create an open loop |
| Preview | 0:15–0:30 | 30–40 words | Promise what the viewer will get |
| Chapter 1 | 0:30–3:30 | ~420 words | First main idea + mini-hook at end |
| Chapter 2 | 3:30–6:30 | ~420 words | Second main idea + mini-hook at end |
| Chapter 3 | 6:30–8:30 | ~280 words | Third idea or case study |
| Recap | 8:30–9:30 | ~140 words | Summarize the 3 key takeaways |
| CTA | 9:30–10:00 | ~70 words | Subscribe ask + next video tease |
English speaking pace is roughly 130–150 words per minute in a conversational YouTube delivery. That means a 10-minute video needs 1,300–1,500 words — not 3,000. If your first ChatGPT draft runs long, ask it to "trim Chapter 2 by 30%, keeping all three examples" rather than cutting the whole section.
Before vs. After: Turning a Weak Intro into a Hook
A generic opening loses most viewers by second 10. Telling ChatGPT to rewrite it as a pattern-interrupt hook — leading with a result, a surprising stat, or a direct challenge — produces intros that stop the scroll. The prompt is a single instruction: "rewrite this intro using the result-first hook pattern, under 20 words."
Here's a real-world comparison. A creator scripted an intro for a productivity video the usual way, then used ChatGPT to rewrite it. The difference in the first sentence is stark:
Hey guys, welcome back to my channel. Today we're going to be talking about productivity apps and how they can help you get more done in your day. I've been using these for a while now and I really think they can make a difference, so let's get into it.
I cleared my entire to-do list by noon for 30 days straight. Three apps made it possible — and none of them cost anything. Here's exactly how I set it up.
The rewritten version is 40 words shorter and front-loads a result the viewer actually wants. Here's the prompt that produced it:
(Role) You are a YouTube scriptwriter who specializes in high-retention hooks. (Context) My channel covers productivity tools for remote workers. Here is my current intro: [paste weak intro]. (Task) Rewrite this as a result-first hook — open with the outcome the viewer wants, then tease the method, all under 40 words. (Format) Give me 3 variants. Label each one by hook type (result-first / question / pattern-interrupt).
Variant 1 (result-first): I cleared my entire to-do list by noon for 30 days straight. Three apps made it possible — and none of them cost anything. Here's exactly how I set it up.
Variant 2 (question): What if your productivity problem isn't motivation — it's your setup? I spent 30 days testing every popular productivity app so you don't have to. Here are the three that actually stuck.
Variant 3 (pattern-interrupt): Stop downloading productivity apps. Seriously. Most of them are making you less productive. I'll prove it in the first 60 seconds — and then show you the three that are different.
6 Copy-Paste ChatGPT Prompts for YouTube Scripts
Six prompts cover every common YouTube script need: a multi-variant hook generator, a full tutorial script with B-roll cues, a reverse-countdown list video, a 60-second Shorts script, a B-roll note injector for existing scripts, and a closing CTA with subscribe ask. Each uses the Role–Context–Task–Format structure with [brackets] for your details.
1. Hook Generator — 5 Variants
2. Full Tutorial Script (Hook → CTA)
3. List Video — Reverse Countdown
4. 60-Second Shorts Script
5. B-Roll Cue Injection (Existing Script)
6. Closing CTA + Subscribe Ask
Prompting for B-Roll Notes, Shorts, and CTAs
B-roll prompts tell ChatGPT to insert [B-ROLL: …] markers after each scripted beat, giving you a shooting list alongside the voiceover. For Shorts, constrain the prompt to 130–150 words and a sub-3-second first sentence. CTA prompts work best when you specify the next video title so the ending feels earned, not bolted on.
The B-roll injection prompt (card 5 above) is especially useful if you've already written or recorded a rough script manually. Drop the transcript in, get a location-specific shot list back in under a minute. Here's what that exchange looks like for a tech review video:
(Role) You are a YouTube Shorts scriptwriter. (Context) Topic: one productivity tip — using a "shutdown ritual" to actually stop working at the end of the day. Channel niche: remote work and work-life balance. Tone: calm, direct, no hype. (Task) Write a 60-second Shorts script. First sentence hooks in under 3 seconds. One idea only. Final sentence is a follow ask. (Format) 130–150 words. Max 12 words per sentence. Add [CAPTION] markers at key moments.
If you can't stop working, try this. [CAPTION: The Shutdown Ritual]
Every day at 5pm, I do three things. I write down every open task. I check tomorrow's calendar. Then I say out loud: "Shutdown complete."
That phrase sounds ridiculous. It works anyway. [CAPTION: Why it works]
Your brain doesn't know when "work" ends if you never signal it. The ritual creates a boundary. A real one.
After two weeks, I stopped checking email after 6pm. Completely. [CAPTION: No more 10pm Slack]
It's not about discipline. It's about designing a clear stop signal.
If this helped, follow for one practical remote work tip every week.
Notice the structure: the first 5 words are the hook, each sentence stays short, and the follow ask is tied to a specific content promise ("one practical tip every week") rather than a generic "follow me." That's the difference between a CTA that converts and one that gets skipped.
One more thing worth knowing: English speaking pace on YouTube typically runs 130–150 words per minute in conversational delivery (faster in energetic content, slower in tutorials). If ChatGPT gives you a 2,000-word script for a 10-minute video, it'll run 13–15 minutes on camera. Ask it to recalculate: "Estimate the speaking time of this script at 140 wpm and trim to hit 10 minutes."
Frequently Asked Questions
What is the best ChatGPT prompt for a YouTube script?
Give ChatGPT your role (YouTube scriptwriter), context (channel niche, target audience, video length), task (write a complete script with Hook → Preview → 3 chapters with mini-hooks → Recap → CTA), and format (short sentences, contractions, [B-ROLL] markers). Generate 3–5 hook variants and select the strongest before you finalize the script.
How long should a YouTube script be for a 10-minute video?
At a conversational delivery pace of 130–150 words per minute, a 10-minute video needs 1,300–1,500 words. Ask ChatGPT to estimate the speaking time of your draft at 140 wpm, then trim by chapter if it runs over. Cutting the recap or CTA to hit a word count is a mistake — trim the body chapters instead.
Can ChatGPT write a Shorts script?
Yes. Constrain the prompt with three rules: the first sentence must be a hook under 3 seconds, the total word count must be 130–150 words, and the script must cover exactly one idea. Final sentence is a follow or like ask. One idea per Short is the non-negotiable rule — trying to cover two points kills watch time on Shorts faster than anything else.
How do I make a ChatGPT YouTube script sound natural on camera?
Add spoken-language constraints to your prompt: sentences under 15 words, contractions throughout, conversational connectors like "but here's the thing" or "and honestly," and rhetorical questions to create micro-pauses. If you have previous video transcripts, paste 2–3 paragraphs as style examples and ask ChatGPT to match your voice. Run the final script through one spoken read-aloud before filming — anything that trips your tongue needs a rewrite.
What is a hook in a YouTube script?
A hook is the first 5–15 seconds designed to prevent the viewer from clicking away. The four most reliable hook patterns are: leading with the end result, posing a question the viewer is already asking, opening mid-action (something is already happening), or making a counterintuitive claim that challenges an assumption. ChatGPT can generate all four variants for any topic in a single prompt — pick the one that fits your channel voice.
Does using AI for YouTube scripts violate any policies?
No. YouTube evaluates content quality and viewer engagement, not the tools used to produce a script. The risk is factual errors — AI occasionally generates plausible-sounding but inaccurate information. Always fact-check numbers, product details, and any claims before filming. Transparency with your audience about AI-assisted production is good practice but is not a YouTube policy requirement.
Putting It Together
The scripting stack comes down to four moves: generate 5 hooks and pick one, use the retention structure as your scaffold, write spoken-language constraints into every prompt, and inject B-roll cues so your script doubles as a shooting list. None of this is complicated — it just requires being specific with ChatGPT instead of vague.
If you want to extend this workflow beyond scripting, the same prompt discipline applies to ChatGPT prompts for marketing — content briefs, ad copy, and campaign strategy all run on the same role-context-task-format logic. The model doesn't change; the domain does.
Comments
Comments (0)
Leave a Comment