ChatGPT Prompts for YouTube Scripts (2026): 20+ Tested Prompts
ChatGPT prompts work best when paired with niche data — feed real outlier scripts, audience pain points, and hook patterns into the prompt, and the output stops sounding like ChatGPT.
This is a library of 20+ ChatGPT prompts for every stage of YouTube script writing — hook generation, body outlines, mid-video and outro CTAs, retention editing and rewrites, and niche-specific adaptation. Each prompt has been tested in ChatGPT and includes a pro tip on how to push the output further. For the broader scripting playbook these prompts plug into, see the YouTube Script Writing guide.
TL;DR
- • ChatGPT scripts only sound AI-generated when the prompt is generic — feed real niche data and the output gets sharper.
- • This library covers 5 prompt categories: hooks, outlines, CTAs, editing, and niche adaptation.
- • Every prompt is copy-paste ready, with a use-case line and a pro tip on how to push the output further.
- • For the underlying script structure these prompts plug into, see the YouTube Script Writing hub.
Key Takeaways
| Prompt Category | Best For | Time Saved | Limitation |
|---|---|---|---|
| Hooks | First 30 seconds | 20-30 min per video | Generic openers without niche data |
| Outlines | Body structure, retention pacing | 30-45 min per video | Misses niche-specific objections |
| CTAs | Mid-video and outro CTAs | 15 min per video | Defaults to clichés without prompting |
| Editing | Trimming, de-AI-ifying, voice rewrites | 45-60 min per video | Cuts personality first if unguarded |
| Niche adaptation | Reusable system prompts, faceless, Shorts | Compounding over many scripts | Needs honest niche context to work |
How to Use These Prompts Effectively
Every prompt below uses placeholders in [BRACKETS]. Those brackets are not optional — they're where you feed the model real niche context. A prompt that says [NICHE] works when you write "cooking creators on a budget"; it doesn't work when you write "food".
ChatGPT does not know what's working in your specific niche this week. It pattern-matches against the general internet. The fastest way to get scripts that read like a niche expert wrote them is to give the model a real expert's context: top-performing video transcripts, exact audience pain points, hook patterns from outlier videos, and a sample of your own voice.
The strongest workflow is: research with data → prompt with that data → edit in your voice. Tools like OutlierKit's Outlier Finder and Script Analysis are designed for the research step — they surface what's actually working in your niche so the prompt has something concrete to anchor on.
The Prompt Library
1. Hook Generation Prompts
The first 30 seconds decide whether viewers stay. Use these prompts to generate hook variants quickly, then test each one out loud before committing.
10-Hook Variant Generator
Use when: You have a topic and title but no hook yet.
You are a YouTube scriptwriter who has studied the top 100 videos in the [NICHE] niche. Topic: [TOPIC] Working title: [TITLE] Target audience: [WHO IS WATCHING] Video length: [LENGTH IN MINUTES] Write 10 hook options for the first 15 seconds of this video. Constraints: - Each hook is 2 sentences max. - No "hey guys", no "in this video", no self-introduction. - Use one of these openers: a result, a counterintuitive claim, a specific pain point, a number, or a direct question. - Label each hook with the opener type used. Return as a numbered list with the opener type in parentheses after the hook.
Pro tip: Read the top 3 hooks aloud. The one that sounds most natural in your voice — not the cleverest on paper — is usually the winner.
Pattern-Match Hook from a Winner
Use when: You found an outlier video and want a hook that matches its energy without copying it.
Here is the opening 30 seconds of a YouTube video that performed well in my niche: """ [PASTE TRANSCRIPT OF FIRST 30 SECONDS] """ 1. Identify the hook structure used (open loop, contrarian claim, before/after, problem statement, etc.). 2. Identify the rhetorical devices (specific numbers, sensory language, curiosity gaps). 3. Write 5 new hooks for my topic "[TOPIC]" using the same structure but with completely different wording. 4. Briefly explain what each hook borrows from the original.
Pro tip: Paste real transcripts from outlier videos in your niche, not your own past scripts. The model needs proven structures, not your average ones.
Curiosity-Gap Hook for Listicles
Use when: Writing a list-style video and need a hook that prevents skip-to-the-end behavior.
Write 5 opening hooks for a YouTube listicle video. Topic: [TOPIC] Number of items: [N] Standout item: item #[X] — [WHY IT'S SURPRISING] Each hook must: - Tease item #[X] without naming it. - Promise that the order is intentional (not random). - Be under 20 words. - Avoid the phrase "you won't believe". Return only the hooks, no explanation.
Pro tip: After picking a hook, add a mid-video reminder around 50% (e.g. "item #X is coming up") — it's the single biggest retention lift on listicles.
Story Hook from Personal Experience
Use when: Writing a case study or personal-journey video.
I'm writing a YouTube video about [TOPIC]. Here is the personal story behind it: """ [3-5 SENTENCES OF YOUR STORY — what happened, when, what changed] """ Write 5 cold-open hooks (first 10 seconds, spoken in first person) that: - Drop the viewer into the most dramatic moment of the story. - End on a tension-building line that makes them need to know what happened next. - Avoid summarizing the whole video. For each hook, explain what tension it creates.
Pro tip: Don't open with the lesson — open with the moment things broke. The lesson is the payoff, not the hook.
2. Body / Outline Prompts
Outlines decide whether a video has a spine. Use these to convert raw notes into a structured body that tracks one promise from hook to CTA.
Promise-Backed Outline
Use when: You know your hook and want the body to deliver on it without drift.
My YouTube video hook is: """ [HOOK] """ The promise to the viewer is: [ONE-SENTENCE PROMISE]. Video length target: [LENGTH] minutes. Audience: [WHO]. Build a section-by-section outline that delivers on that promise. Constraints: - 4-7 sections total, each 60-180 seconds. - For each section, give: section title, the single point being made, the proof (data, example, demo) used, and a transition line to the next section. - Cut anything that doesn't directly support the hook's promise. - Flag any section that risks audience drop-off and explain why.
Pro tip: If a section can't be written in one sentence, it's two sections. Force one-sentence summaries before writing prose.
Tutorial Step-by-Step Outline
Use when: Writing a how-to or tutorial video.
I am writing a tutorial video on: [TASK]. Skill level of viewer: [BEGINNER / INTERMEDIATE / ADVANCED]. Tools or context the viewer already has: [LIST]. Produce a tutorial outline with: 1. A 30-second context bridge after the hook. 2. Numbered steps (each step = one screen action or one decision). 3. For each step: the action, why it matters, the most common mistake, and a 1-sentence on-screen text overlay. 4. A "what to do if this doesn't work" troubleshooting block. 5. A closing recap with a single next action. Keep each step under 90 seconds of speaking time (~200 words).
Pro tip: Add the troubleshooting block before the recap, not after. It catches viewers who would otherwise drop off frustrated.
Retention-Curve Outline (Pattern Interrupts)
Use when: Your videos drop off mid-way and you want a structure that keeps viewers past 50%.
Here is my draft outline: """ [PASTE OUTLINE] """ Audit it for retention risk: 1. Mark every point where I go more than 90 seconds without a pattern interrupt. 2. Suggest a specific pattern interrupt at each mark (cut to B-roll, on-screen stat, direct question to viewer, format switch, callback to the hook). 3. Identify any section that explains *what* without ever showing *how* — those are usually drop-off zones. 4. Suggest one mid-video tease (around the 40-50% mark) that pulls viewers to the end.
Pro tip: ChatGPT will over-suggest pattern interrupts. Keep only the ones tied to a real visual or example you can produce.
Outline-from-Bullets Cleanup
Use when: You have raw bullet notes and need a coherent script outline.
Here are my raw notes for a YouTube video: """ [PASTE BULLETS] """ 1. Group these notes into 3-6 thematic sections. 2. Order the sections so each one earns the next (cause → effect, problem → solution, simple → complex). 3. For each section, write: the section heading, the single most important point, and the strongest example or stat from my notes. 4. Flag any bullet that doesn't fit and explain why I should cut it. 5. Suggest one section I'm missing that the audience would expect.
Pro tip: Trust the "flag what to cut" output more than the "what's missing" output. Cuts are easier to verify than additions.
3. CTA and Retention Prompts
Most YouTube CTAs are forgettable because they're generic. These prompts produce CTAs tied to the specific value the viewer just received.
Mid-Video Engagement CTA
Use when: You want a comment-driving CTA at the natural pause around 40-50%.
My video is about: [TOPIC]. The audience is: [WHO]. At the 50% mark, I want to ask one question that gets thoughtful comments. Write 5 mid-video CTAs that: - Ask one specific question (not "what do you think?"). - Tie directly to the section the viewer just watched. - Give 2-3 example answer formats so people don't have to think from scratch. - Stay under 25 words each. For each, explain what kind of comments it will produce.
Pro tip: Pick the CTA whose example answers you would actually want to read. That's the one your audience will too.
End-of-Video Subscribe CTA
Use when: Writing the outro CTA that drives subscribes (not just views).
My channel is about: [CHANNEL TOPIC]. The video the viewer just watched: [THIS VIDEO TOPIC]. The specific transformation a subscriber gets over time: [WHAT THEY LEARN OR ACHIEVE]. Write 5 outro subscribe CTAs that: - Reference what they just watched (not "if you liked this video"). - Promise a specific recurring value, not "more videos like this". - Mention one upcoming or related video by name. - Are 30-45 seconds when read aloud (~80-100 words). Avoid: "smash the like button", "ring the bell", "comment below".
Pro tip: The strongest subscribe CTA names a problem the viewer hasn't solved yet and points to the next video that solves it.
End-Screen Next-Video Pitch
Use when: You need a 20-second pitch for the next video that fills the end screen.
The video I want to send viewers to next is: [NEXT VIDEO TITLE]. That video covers: [ONE-LINE SUMMARY]. The viewer just watched a video about: [CURRENT VIDEO TOPIC]. Write a 15-20 second end-screen pitch that: - Names the specific problem the next video solves. - Connects it to what they just learned. - Ends 5 seconds before the end screen so the cards display silently. - Avoids the words "check out", "go watch", "I made another video". Return 3 versions in different tones: matter-of-fact, urgent, and curiosity-based.
Pro tip: Match the tone version to your channel's voice. "Curiosity" works for entertainment, "matter-of-fact" works for educators.
Retention Rescue: Drop-off Diagnosis
Use when: You've published the video and YouTube Studio shows a sharp drop-off you want to fix in the next script.
Here is the script section that lost the most viewers in my last video (drop from [X]% to [Y]% retention): """ [PASTE SCRIPT EXCERPT] """ Audience: [WHO]. Video topic: [TOPIC]. 1. List the 3 most likely reasons viewers dropped off here (be specific to the words used, not generic). 2. Rewrite this section in 3 different ways that fix those reasons. 3. For each rewrite, explain which retention failure it addresses. 4. Suggest where in the section a pattern interrupt would have helped.
Pro tip: Always paste the actual script, not a paraphrase. The model needs the real phrasing to spot what made viewers leave.
4. Script Editing and Rewriting Prompts
Editing is where AI saves the most time. These prompts tighten pacing, remove AI-isms, and convert written prose into spoken voice.
Cut 20% Without Losing Substance
Use when: Your draft is too long and you can't see what to cut.
Here is my YouTube script draft:
"""
[PASTE DRAFT]
"""
Cut this script by 20% without losing any substantive point.
Rules:
- Remove filler phrases ("basically", "essentially", "what I mean is").
- Collapse any sentence that restates the previous one.
- Cut adjectives that don't change meaning.
- Keep every example, stat and story beat.
- Mark each cut with [CUT: reason] so I can review.
Return the trimmed script and a count of words removed.Pro tip: Review the [CUT] markers — accept the obvious filler cuts, push back on cuts that remove voice. ChatGPT trims personality first.
De-AI-ify a Script
Use when: Your script reads like ChatGPT wrote it.
Here is a YouTube script that sounds AI-generated: """ [PASTE SCRIPT] """ Rewrite it so it sounds like a human creator speaking, specifically: - Replace all "delve", "dive into", "in today's video", "let's explore", "the world of", "moreover", "furthermore", "in conclusion". - Use contractions everywhere (I'm, you'll, don't, here's). - Replace abstract nouns with concrete examples (instead of "engagement strategies" → "the question I ask at minute 4"). - Break sentences over 25 words into 2-3 short ones. - Keep all factual content and structure. Return the rewrite and a 5-bullet list of the patterns you replaced most.
Pro tip: Maintain a personal stop-list of words you never say out loud. Add them to this prompt for your own voice.
Convert Written Prose to Spoken Voice
Use when: You wrote the script as an essay and now it doesn't flow when read aloud.
Here is a section written in essay style: """ [PASTE PROSE] """ Rewrite it for spoken delivery on YouTube: - Use sentence fragments where natural. - Add one rhetorical question every 200 words to break monotony. - Replace any sentence I would not actually say out loud. - Mark spots for breath / pause with [BEAT]. - Mark spots that need a B-roll or visual with [B-ROLL: description]. Read it back to me as a script with stage directions inline.
Pro tip: Read the output aloud against a stopwatch. If your speaking pace doesn't match the pacing the model assumed, ask for a faster or slower rewrite.
Hook-to-Outro Coherence Check
Use when: You've finished a draft and want to make sure the script keeps its promise.
Here is my full draft: """ [PASTE FULL DRAFT] """ Audit the script for promise-coherence: 1. State the promise the hook makes (in your own words). 2. List every section and rate (1-5) how directly it delivers on that promise. 3. Identify the single biggest digression and explain its cost. 4. Confirm whether the outro CTA pays off the hook's promise or introduces a new one. 5. Give one structural change that would tighten the through-line.
Pro tip: If the model's restated promise doesn't match what you intended, your hook is the problem — fix it before editing the body.
5. Niche-Specific Adaptation Prompts
Generic prompts produce generic scripts. These adapt the prompt itself to your niche, audience and competitors so output is specific from the first generation.
Build a Custom Niche System Prompt
Use when: You'll be using ChatGPT for many scripts in the same niche and want better defaults.
I want you to act as a YouTube scriptwriter specialized in [NICHE]. Use the following permanent context for every script we write together: - Audience: [DEMOGRAPHICS, EXPERIENCE LEVEL, MOTIVATIONS] - Audience pain points: [3-5 SPECIFIC PAIN POINTS] - Audience language patterns: [PHRASES THEY USE, PHRASES THEY DON'T] - My channel's voice: [3 ADJECTIVES + 1 EXAMPLE LINE I'D ACTUALLY SAY] - Forbidden phrases: [LIST] - Top 3 outlier videos in this niche right now: [TITLES + ONE-LINE WHY THEY WORKED] Confirm you understand by: 1. Restating the audience in one paragraph. 2. Writing one sample hook that proves you got the voice. 3. Listing 5 angles I haven't covered that fit this audience. Then wait for the topic of the first script.
Pro tip: Save this conversation and reuse the same chat for every script in this niche. The cumulative context becomes your strongest asset.
Faceless Channel Script Adapter
Use when: Adapting a personality-led script style for a faceless or voiceover channel.
Here is a script written for an on-camera creator: """ [PASTE SCRIPT] """ Adapt it for a faceless YouTube channel where: - All visuals are stock footage, screen recordings, or AI-generated. - The narrator is a single voice with no on-screen presence. - The audience expects information density over personality. Changes to make: - Replace any reference to facial expression, body language, or "look at this". - Replace personal anecdotes with sourced facts or examples. - Add visual cue notes [VISUAL: ...] where the original relied on the creator's presence. - Tighten pacing — faceless audiences tolerate less filler. Return the adapted script with visual cues inline.
Pro tip: Faceless scripts live or die by the visual cues. Spend more editing time on [VISUAL] notes than on prose.
Shorts Script from Long-Form
Use when: You want to repurpose a long-form video into a 45-second Short.
Here is the strongest 60-second segment of my long-form video: """ [PASTE TRANSCRIPT] """ Convert it into a 45-second YouTube Short script: - Hook in the first 2 seconds (max 8 words). - Vertical-friendly visual structure (one idea per cut). - One clear takeaway, not three. - A loop-back final line that makes viewers rewatch. - A single call to action: watch the full video, with a one-line tease of what they'll get. Return: the script, the on-screen text overlays, and 3 alternative hooks.
Pro tip: Test the 3 alternative hooks as 3 separate Shorts. Hook variance matters more than full-script variance on Shorts.
Localize a Script for a New Audience
Use when: Adapting a working script for a different region, language register, or audience segment.
Here is a script that worked for [ORIGINAL AUDIENCE]: """ [PASTE SCRIPT] """ Adapt it for [NEW AUDIENCE] where: - Cultural references different: [ORIGINAL → REPLACE WITH] - Currency / units / examples: [ORIGINAL → REPLACE WITH] - Tone shift: [e.g., more formal, more casual, more direct] - Forbidden references / sensitivities: [LIST] Rules: - Keep the structure and hook strategy. - Replace every specific example with one resonant for the new audience. - Flag anything you couldn't confidently localize and explain why. Return the adapted script and a list of every substitution you made.
Pro tip: Always review the substitution list manually. Cultural localization is where the model is most likely to confidently invent.
Why Prompts Alone Aren't Enough
ChatGPT is a structure machine. Give it a clear scaffold and a few constraints and it will produce a plausible YouTube script every time. That's the trap — plausible isn't the same as performative. The scripts that actually win on YouTube are specific to a niche, a moment, and an audience that has heard every generic hook a hundred times already.
ChatGPT doesn't know what's outlying in your niche this week. It doesn't know which thumbnails are working, which hook structures the algorithm is currently rewarding, or which audience pain points have moved since its training cutoff. When you prompt without that context, the model fills the gap with averages — and averages don't outlier.
The strongest workflow is a loop, not a single prompt: research what's working → prompt ChatGPT with that data → edit in your voice → measure → feed real retention data back into the next prompt. The research step is where most creators skip — and it's also where the leverage lives.
That's the gap OutlierKit fills. Outlier Finder surfaces the videos in your niche that are 3-10x outperforming the channel average — those transcripts, hooks and structures are what you paste into the "Pattern-Match Hook from a Winner" prompt above. The two together turn a blank ChatGPT box from a structure generator into a niche-aware co-writer.
Then score the draft before you record
Once ChatGPT has produced a draft, run it through OutlierKit's Video Analyzer. It breaks down a YouTube script and scores it across the seven dimensions that decide whether a script performs:
- Hook strength
- Curiosity loops
- Pacing & structure
- Emotional triggers
- Pattern interrupts
- Promise delivery
- Algorithm optimization
The output tells you which dimensions the script nails and which ones need a rewrite — applied to your draft, not a generic checklist. The smartest workflow is: run Video Analyzer on the top outlier in your niche to see what a 9/10 script looks like, then run it on your ChatGPT draft and feed the gap back into the "Retention Rescue" or "Hook-to-Outro Coherence Check" prompts above.

OutlierKit Video Analyzer — live demo
That loop — outlier research → prompt with that data → score the draft → rewrite the weak dimensions — is what separates a generic ChatGPT script from one that actually retains.
You don't need OutlierKit to use these prompts — they work fine on their own. But you'll feel the ceiling fast if the [PASTE TRANSCRIPT] field is always blank and the only feedback on your draft is your own re-read.
Frequently Asked Questions
Basics
Does ChatGPT write good YouTube scripts?
ChatGPT writes good first drafts and great hooks once you give it real niche context. It cannot independently know what's working in your specific niche this week, so a script generated from a one-line prompt will read generic. The reliable workflow is: research outlier videos, paste real performance context into the prompt, let ChatGPT structure the script, then edit in your voice and examples.
Free ChatGPT vs ChatGPT Plus for scripts — which do I need?
For occasional script drafts, the free tier is enough. ChatGPT Plus ($20/month at the time of writing) is worth it if you write more than one script per week, want longer outputs without truncation, or want to use Custom GPTs to lock a niche-specific system prompt. Verify current pricing on openai.com before subscribing — OpenAI changes tiers regularly.
Workflow
What's the best ChatGPT model for YouTube scripts?
Use the most capable general model available in your account for outlines and rewrites, since coherence over long context matters more than raw speed. For quick hook variants and short edits, faster smaller models are fine. Test on a script you already know performs well — whichever model preserves your voice with fewer corrections is the right one for your workflow.
How do I make ChatGPT scripts not sound AI-generated?
Three things in combination: (1) feed ChatGPT a sample of your own writing or a real transcript from your channel and ask it to match the voice, (2) maintain a stop-list of AI tells ("delve", "in today's video", "let's explore") and explicitly forbid them in the prompt, (3) always do a final read-aloud pass — anything you wouldn't actually say gets rewritten manually.
Should I use Custom GPTs for YouTube scripts?
Yes, if you write in one niche consistently. A Custom GPT lets you save your audience profile, voice rules, forbidden phrases, and outlier examples so every new script starts with full context. The upfront setup takes 30-60 minutes but compounds over every future script. The "Build a Custom Niche System Prompt" template above is a starting point.
Limitations
Why do my ChatGPT scripts get low retention?
The most common reason is that the script is structurally fine but generically argued — the examples, stats and stories aren't from your niche. ChatGPT pads with abstract claims when it lacks real context. Fix it by pasting actual outlier transcripts, real audience comments, or your own past performance data into the prompt before generating. Structure is the cheap part; specificity is what holds attention.
Can ChatGPT analyze competitor scripts directly?
ChatGPT can analyze a competitor script if you paste the transcript in. It cannot pull a video, watch it, and analyze it on its own. For repeatable competitor analysis tied to actual view performance and outlier signals, use a dedicated tool like OutlierKit's Script Analysis, then feed the findings into ChatGPT for the writing step.
Are these prompts safe to use commercially?
The prompts themselves are templates — you can use them in any commercial workflow. Output from ChatGPT is governed by OpenAI's terms of use, which currently allow commercial use of generated content for paid users. Verify terms on openai.com before relying on this for client work, since policies change.
Related Resources
Ready to grow your YouTube channel?
OutlierKit helps you find winning content strategies with competitor analysis and keyword research.
Try OutlierKit Free