23 Claude Prompts for YouTube Creators (2026)
The complete prompt library for research, hooks, scripting, and titles, built around how the 2026 YouTube algorithm actually evaluates content.
87% of creators now use AI in their workflows. Most of them type "write me a YouTube script about X" and wonder why the output sounds like a corporate press release.
The problem isn't Claude. The problem is the prompt.
Claude is a reasoning engine. Give it context, constraints, and a specific output format and it produces work that a professional writer would respect. Give it a vague one-liner and you get something your audience will click away from in 8 seconds.
This guide is the complete library of Claude prompts for YouTube creators that I wish existed when I started. 23 prompts across six categories: research, hooks, scripting, titles, SEO, and a workflow note at the end. You can copy-paste any of these today.
Why Claude Prompts for YouTube Creators Usually Fail
Most YouTube creators who've tried Claude for content have experienced the same thing: the output is technically correct but emotionally flat. It passes a grammar check. It fails the viewer test.
The reason is almost always the same. The prompt has no context.
Claude doesn't know your niche. It doesn't know your audience's specific frustration level. It doesn't know whether your viewers are 22-year-old side hustlers or 45-year-old finance professionals trying to retire early. Without that information, it defaults to the median, and the median is forgettable.
The prompts in this guide are built around three principles that separate working prompts from wasted prompts:
- Context before request. Tell Claude who your audience is, not just what the video is about.
- Format constraint. Specify the output structure you need, not just the content.
- Evaluation criteria. Tell Claude how to judge what it writes, not just what to write.
Every prompt below follows these principles. The difference in output quality is immediate and measurable.
Claude Prompts for YouTube Research: Find the Angle Before You Write a Word
The 30 seconds after a viewer clicks matters more than anything else in your video. But before you can write a hook that holds attention, you need to know exactly what your audience already believes, and where those beliefs have gaps.
These five prompts handle that research layer.
Prompt 1: The Outlier Angle Finder
Most niches have three or four things every creator says. This prompt finds the counterintuitive crack in the conventional wisdom.
I'm making a YouTube video about [TOPIC]. Before I write a single word, I need to find the counterintuitive angle.
Here is what I know most creators say about this topic:
- [COMMON CLAIM 1]
- [COMMON CLAIM 2]
- [COMMON CLAIM 3]
I need you to identify:
1. The one claim above that is actually incomplete or misleading, and what the more complete truth is
2. A counterintuitive fact inside this topic that most creators ignore because it contradicts the popular narrative
3. The specific audience belief this counterintuitive truth would challenge
Write each finding in one sentence. Do not write explanations yet. Just the three claims.Why it works: Claude has read more niche content than you have. Asking it to find the crack in the consensus surfaces angles your competitors haven't taken.
Prompt 2: Comment Mining for Unspoken Pain Points
YouTube comments are the most underused research source in the creator economy. The comments on your competitors' videos are a direct transcript of what your audience still wants to know.
Below are 20 YouTube comments from videos about [YOUR TOPIC]. I pulled them from the top 3 videos in my niche.
[PASTE COMMENTS HERE]
From these comments, identify:
1. The most common unresolved frustration (what people are still stuck on after watching)
2. The question that keeps appearing in different forms (the thing viewers want answered that the original video didn't fully address)
3. One comment that reveals a belief the audience holds that is factually incorrect
Format the output as three bullet points, each starting with the category name in brackets: [FRUSTRATION], [RECURRING QUESTION], [WRONG BELIEF].Why it works: You're not asking Claude to guess what your audience wants. You're feeding it real signal and asking it to find the pattern. The wrong-belief finding is particularly powerful for hook writing.
Prompt 3: Niche Gap Analyst
This prompt identifies the content no one in your niche is making. Before you start it, collect the titles and descriptions of the 10 most-viewed videos on your topic from the past six months.
Here are the titles and descriptions of the 10 most-viewed videos on [YOUR TOPIC] from the past 6 months:
[PASTE TITLES AND DESCRIPTIONS HERE]
Analyze this content for:
1. The subtopics every video in this list covers (the saturated ground)
2. The subtopics no video in this list covers (the gap)
3. The specific audience segment that none of these videos appears to be written for
Do not recommend video ideas yet. Just surface the gaps in three clearly labeled sections.Why it works: Content gaps are easier to find systematically than intuitively. Claude can pattern-match across 10 data points in seconds. What takes a creator 45 minutes of tab-switching takes Claude about 90 seconds.
Prompt 4: Trend Velocity Tester
Not every topic deserves a video right now. This prompt helps you assess whether a topic is growing, peaking, or fading before you invest production time.
I want to make a video about [TOPIC]. This topic has been discussed in my niche for approximately [X months or years].
Based on your training data, tell me:
1. Is the search interest for this topic growing, stable, or declining?
2. What is the likely peak interest window? When is this topic most likely to be searched?
3. What is the risk of this topic becoming outdated or oversaturated in the next 6 months?
Rate each answer on a 1-10 confidence scale and explain what would change your assessment.Why it works: Claude won't give you live Google Trends data, but it can assess trajectory based on its training corpus. The confidence score forces it to flag where it's uncertain, which tells you where to do additional research.
Prompt 5: The Idea Validator
Use this before you commit to filming. Three ideas go in. One clear winner comes out.
I have three potential YouTube video ideas for my channel. My channel is about [NICHE]. My audience is [DESCRIBE: age range, primary goal, biggest current frustration].
Here are the three ideas:
1. [IDEA 1]
2. [IDEA 2]
3. [IDEA 3]
For each idea, score it on:
- Search demand: will people be actively searching for this? (1-10)
- Audience specificity: does this speak to MY audience vs. everyone? (1-10)
- Differentiation: is there a clear angle separating this from existing content? (1-10)
Give a composite score for each and a one-sentence verdict on which to make first and why.Why it works: The three-score framework forces Claude to evaluate ideas separately on different axes. Most creators make the mistake of picking ideas based on search volume alone. This prompt corrects that.
Hook Prompts: The First 30 Seconds That Determine Everything
71% of viewers decide whether to keep watching within the first 3 seconds. 55% are gone by the 60-second mark if the introduction hasn't delivered something that makes them feel the video is worth their time.
This is not a content quality problem. It is a hook architecture problem. The prompts below fix it.
Prompt 6: Pattern Interrupt Hook Generator
A pattern interrupt breaks the expected opening. Instead of "Today I'm going to show you," the video opens with a statement that stops the viewer mid-scroll.
I'm writing the first 30 seconds of a YouTube video about [TOPIC].
My audience's biggest assumption about this topic is: [STATE THE COMMON BELIEF]
Write 5 different opening lines that immediately break that assumption. Each opening should:
- Be under 15 words
- Contain a specific number or concrete detail
- Create a question in the viewer's mind that only watching the video will answer
Do not explain the hooks. Just list the 5 lines.Why it works: The constraint of 15 words forces Claude out of explanatory mode. The specificity requirement prevents vague openers. You get five options and pick the one that feels most true to your voice.
Prompt 7: Curiosity Gap Hook Builder
A curiosity gap hook withholds just enough information to make not-watching feel like a loss.
I need to open my video about [TOPIC] with a curiosity gap hook.
The payoff the viewer will get by the end of the video is: [STATE THE MAIN INSIGHT OR OUTCOME THEY WILL LEARN]
Write a hook that:
1. Hints at the payoff without revealing it
2. Creates a specific knowledge gap the viewer feels immediately
3. Uses a concrete detail rather than a vague promise
Give me 3 versions: one that leads with a question, one that leads with a statement, and one that leads with a counterintuitive fact.Why it works: Giving Claude the payoff first forces backward-engineering. The hook becomes a promise, not just a teaser.
Prompt 8: Story Hook Builder
Stories hold attention because the brain treats narrative gaps the same way it treats physical gaps. It wants them closed.
I want to open my video about [TOPIC] with a story hook.
The specific moment I want to center the story on: [DESCRIBE A TURNING POINT OR MOMENT OF REALIZATION]
Write a 4-sentence story hook that:
- Opens with the most compelling sentence of the story, not the beginning of the story
- Creates immediate tension or stakes in sentence two
- Gives the viewer exactly one line of context to understand what's happening
- Ends with a cliffhanger that connects directly to the video's main promise
Do not exceed 4 sentences under any circumstances.Why it works: Most story hooks fail because they start at the beginning. This prompt forces Claude to open in the middle, which is where the tension lives.
Prompt 9: Data-Led Hook
A single well-chosen statistic can hold a viewer better than a 60-second story. The trick is the sentence that follows it.
Write me a cold open for a YouTube video about [TOPIC].
The data point I want to lead with: [STATE THE STATISTIC OR FINDING YOU WANT TO USE]
The hook should:
- Open with the data point, stated plainly with no warm-up sentence before it
- Follow with one sentence that makes the viewer feel the personal implication of that number
- End with a one-sentence promise of what this video will deliver
Total length: 3 sentences. No exceptions.Why it works: The three-sentence constraint forces compression. The "personal implication" instruction prevents Claude from just restating the stat in different words.
Prompt 10: Hook Quality Evaluator
Use this on hooks you've already written. It's the fastest way to find the weak point before you film.
Here is the current hook I've written for my YouTube video about [TOPIC]:
[PASTE YOUR HOOK]
Evaluate it against these four criteria:
1. Does it open with something specific (a number, a name, a moment) rather than a vague statement?
2. Does it create a knowledge gap that only the video can close?
3. Does it make an implicit promise that matches what the video actually delivers?
4. Would a viewer who knows nothing about this topic still feel pulled to keep watching?
Score each criterion 1-10. Then give one specific rewrite suggestion for the criterion with the lowest score only.Why it works: Asking for one rewrite instead of four forces Claude to prioritize. You get the most important fix, not a list of changes that creates a new draft problem.
Script Architecture Prompts: Build a Video People Actually Finish
Most videos lose the audience in the middle, not the beginning. The hook might be strong. The ending might be clear. But the 3-minute section in the middle where the creator "explains the thing" is where 40-60% of the remaining audience exits.
The next six prompts address script structure, pacing, retention, and the cuts that most creators are afraid to make.
Prompt 11: Full Script Architect
Use this before you write a single line of script. It builds the architecture first.
I need to write a [LENGTH]-minute YouTube video about [TOPIC].
My audience: [DESCRIBE who they are, what they already know, what outcome they want]
The core insight I'm delivering: [ONE SENTENCE - the main thing they will walk away knowing]
The transformation: [Where they are now] to [where they will be after watching]
Create a scene-by-scene script outline with these labeled sections:
- Hook (0:00-0:30): the opening statement and the tension it creates
- Problem (0:30-2:00): what is broken and why it matters to this specific audience
- Mechanism (2:00-X:XX): how the thing actually works, with individual points titled as outcomes
- Proof (X:XX-X:XX): the evidence, demonstration, or case study
- Resolution (X:XX-end): what to do now and the call to action
For each section, write one sentence describing the emotional state you want the viewer to be in while watching it.Why it works: The emotional state instruction is the key. It forces Claude to think about how the video feels, not just what it covers. Feeling is what makes viewers finish.
Prompt 12: Retention Checkpoint Generator
YouTube Analytics shows you where viewers leave. This prompt adds proactive re-engagement points before they do.
Here is the outline for my YouTube video about [TOPIC]:
[PASTE YOUR OUTLINE]
YouTube data shows 20% of viewers leave in the first 10 seconds and another 35% are gone by the 60-second mark if the introduction does not deliver something meaningful.
Review my outline and identify:
1. Where the first "valley" appears (the section most likely to cause a viewer to pause or leave)
2. A specific re-engagement technique to insert at that point: a new question, a surprising stat, or a tease of what is coming next
3. Two additional points in the video where attention may dip, and a brief note on what to insert at each
Do not rewrite the outline. Add the three interventions as labeled insertions: [RETENTION CHECKPOINT 1], [RETENTION CHECKPOINT 2], [RETENTION CHECKPOINT 3].Why it works: Treating retention like a structural problem rather than a content quality problem changes what you fix. The checkpoint insertions take 2 minutes to add and can move average view duration by meaningful amounts.
Prompt 13: The Pacing Auditor
Paste your draft script here. This prompt reads it the way a viewer would.
Here is my YouTube script for a video about [TOPIC]:
[PASTE YOUR SCRIPT]
Read this as a viewer who has no prior knowledge of this subject. Then identify:
1. The single point where a viewer is most likely to click away, and the specific reason why
2. One section that explains more than it needs to (where the script slows without adding value)
3. The moment where the main payoff arrives. Is it earlier or later than it should be?
Give me three specific edits: one cut, one addition, one reorder. Write out the actual replacement text for each, not just the instruction.Why it works: Asking for the actual replacement text instead of instructions forces Claude to commit. "This section is too long" is useless. "Replace lines 3-7 with this one sentence" is actionable.
Prompt 14: Emotional Rhythm Mapper
This prompt maps the emotional arc of your video before you write it. Most creators plan content. The best creators plan feeling.
I'm writing a script about [TOPIC] for a YouTube audience that [DESCRIBE THEIR EMOTIONAL STARTING STATE: e.g., "is frustrated with slow results and has already tried the obvious approaches" or "is excited but overwhelmed by conflicting advice"].
Map the emotional journey this script should take the viewer on:
- What emotion should they feel in the first 30 seconds?
- Where should their frustration or tension peak?
- Where should they feel the first moment of genuine hope?
- Where should they feel confident that they can actually do this?
- What should be the final emotional note just before the call to action?
Then identify the three most important sentences in a script that would carry the emotional weight at each transition point. Write those three sentences.Why it works: The three sentences exercise reveals which moments in the video need the most craft. Everything else can be workmanlike. These three sentences have to land.
Prompt 15: CTA Optimizer
The call to action is the last thing the viewer hears. Most creators waste it.
Here is the call to action from my YouTube video about [TOPIC]:
[PASTE YOUR CURRENT CTA]
My primary goal for this video is: [SUBSCRIBE / WATCH NEXT VIDEO / CLICK LINK IN DESCRIPTION / LEAVE A COMMENT]
Rewrite my CTA so that:
1. It gives the viewer a specific reason to take action, not just "subscribe if you liked this"
2. It connects the action directly to the value they just received from the video
3. It teases what they will get from taking the action (the next video, the resource, the outcome)
Keep the rewritten CTA under 45 seconds of spoken audio. Match the voice and register of the script I pasted.Why it works: The 45-second audio constraint converts to roughly 110 spoken words. That is enough. Most CTAs run long because creators are nervous about asking. The constraint solves that.
Prompt 16: The Anti-Padding Pass
Every script has padding. This prompt finds it.
Here is my YouTube script:
[PASTE YOUR SCRIPT]
Find every sentence that does one of the following:
1. Restates something already said earlier in the script
2. Uses an adjective where a number or specific example would work better
3. Serves as a transition without adding information ("So now let's talk about...")
4. Hedges instead of commits ("This might work for some people" vs "This works if you do X")
List each offending sentence. For each one, either write a tighter replacement or mark it [CUT].Why it works: Creators over-write because they are afraid of the video feeling "too short." Most videos are exactly the right length after padding is removed. The [CUT] option gives Claude permission to be ruthless.
Title and Thumbnail Copy Prompts: The CTR Gap Is a Packaging Problem
The platform-wide average CTR for YouTube sits at 4-6%. Well-optimized content targeting YouTube Search hits 8-15%. That gap is not explained by video quality. It is almost entirely explained by title and thumbnail decisions.
A video that earns a high click-through rate but loses viewers in the first 15 seconds is actively demoted by the 2026 algorithm. A video with a 6% CTR and 65% retention in the first 30 seconds gets pushed. The packaging and the content have to work together.
Prompt 17: Title Formula Generator
This prompt generates eight title options across four proven formulas. You pick the one that fits.
My YouTube video is about: [DESCRIBE THE TOPIC IN ONE SENTENCE]
The key counterintuitive finding or insight: [STATE THE SURPRISING CLAIM]
My audience's biggest fear related to this topic: [STATE THE FEAR]
The specific result they will get from watching: [DESCRIBE THE OUTCOME]
Write 8 title options using these formats exactly:
- 2 titles using a failure/success ratio ("Only X% of [audience] [achieve goal]")
- 2 titles using a counterintuitive claim ("The [common behavior] is killing your [outcome]")
- 2 titles using a specific number promise ("X [things] that [specific outcome]")
- 2 titles using a year plus a specific claim ("In 2026, [specific change] means [implication]")
Rules: no vague adjectives anywhere. Every title must contain a specific number or measurable claim. No em dashes.Why it works: The four-formula structure prevents Claude from defaulting to the same title pattern repeatedly. The "no vague adjectives" rule produces titles that feel researched rather than written.
Prompt 18: A/B Title Tester
Use this when you have two titles and need a data-informed way to choose.
I have two potential titles for a YouTube video:
Option A: [TITLE A]
Option B: [TITLE B]
My channel niche is [NICHE]. My typical viewer is [DESCRIBE: who they are, what they want].
Evaluate each title on:
1. Click intent: does it create a strong pull to click? (1-10)
2. Keyword alignment: does it match how people actually search for this topic? (1-10)
3. Promise clarity: is the value of watching immediately obvious? (1-10)
Give a composite score for each, a one-sentence verdict, and then write one hybrid title that takes the strongest element from each option.Why it works: The hybrid title instruction is the most valuable output. It forces synthesis rather than just evaluation.
Prompt 19: Thumbnail Text Optimizer
Most thumbnail text fails at 120 pixels wide. That is the size of a mobile YouTube thumbnail. If your text is not readable at that scale, it is invisible.
My video is about [TOPIC]. My thumbnail will show [DESCRIBE THE IMAGE: e.g., "me with a surprised expression standing in front of a whiteboard"].
I need text to overlay on the thumbnail. The text must:
- Be readable at 120px width on a mobile screen
- Use no more than 5 words
- Create curiosity or tension without revealing the answer the video delivers
- Not duplicate the title word-for-word
Generate 5 thumbnail text options that work with the image I described. For each one, write one sentence explaining which psychological trigger it uses (curiosity gap, fear, specificity, social proof, or counterintuitive claim).Why it works: The psychological trigger explanation teaches you the mechanism so you can make better thumbnail decisions without Claude next time.
Prompt 20: Title Critique and Upgrade
Use this on any title you are considering. The critique often produces the best title.
Here is a title I am considering for my YouTube video:
[YOUR TITLE]
Tell me:
1. Is there a vague word in this title that could be replaced with a specific number or outcome?
2. Does the title make a promise, or does it just describe a topic?
3. Is there a more counterintuitive framing for this same subject?
Then write three upgraded versions of this title, each using a different technique from your critique. Label them [VERSION A], [VERSION B], [VERSION C].Why it works: Three upgraded versions is a better output than one rewrite. You can test all three or combine elements from each.
SEO and Description Prompts: The Metadata Most Creators Ignore
YouTube descriptions and chapter titles are metadata, not content. Most creators treat them as an afterthought. The creators in the top 5% of search rankings treat them as a second SEO layer that compounds over time.
Prompt 21: Description Architect
The first two sentences of your description appear in YouTube search results. Everything after that is for the viewer who already clicked.
My YouTube video is about [TOPIC]. The primary keyword I am targeting is [KEYWORD].
Write a YouTube description that:
- Opens with the first 2 sentences containing the primary keyword naturally (these appear in search results before the "Show more" cutoff)
- Includes a 3-5 sentence summary of what the video covers
- Lists 3-5 specific things the viewer will learn, written as outcomes they will achieve, not topics that will be covered
- Ends with a question that invites comments
- Stays under 500 words total
List hashtags at the very end only, never in the body of the description.Prompt 22: Chapter Timestamp Creator
Chapters improve watch time because they give viewers a way to navigate back into the video. They also get indexed by YouTube search.
Here is the full transcript or detailed outline of my YouTube video:
[PASTE YOUR TRANSCRIPT OR OUTLINE]
Create YouTube chapter timestamps that:
- Start with "0:00 Introduction" as required by YouTube
- Name each chapter as an outcome or insight, not a topic label ("Why 90% of hooks fail" not "Section 2: Hooks")
- Space each chapter at least 90 seconds apart
- Cover every major section without exceeding 8 chapters total
Return the timestamps in the exact format YouTube expects:
0:00 Chapter NamePrompt 23: Keyword and Tag Strategy
This prompt generates a full keyword architecture for a single video, not just a tag list.
My video is about [TOPIC]. My channel covers [BROAD NICHE].
Generate a keyword strategy with:
1. The primary keyword: what people type into YouTube Search to find this exact video
2. 5 secondary keywords: related searches that would lead an interested viewer to this content
3. 5 broad channel-level tags: niche keywords that describe the category of your channel
4. 3 long-tail phrases: specific questions your target viewer is likely searching word-for-word
For each keyword and phrase, label it as one of: High Volume / High Competition, Medium Volume / Medium Competition, or Long-Tail / Lower Competition.The Workflow Problem Nobody Talks About
Here is the honest part of this guide.
These 23 prompts work. I use versions of all of them. The output quality compared to generic prompts is not subtle, it is immediate.
But there is a problem with running them the way I just described.
Each prompt lives in isolation. You research in one Claude conversation. You script in another. You validate your title in a third. You check your analytics in a dashboard that has no memory of what you scripted last week. Then you repeat the whole loop for the next video with zero accumulated context from what actually worked.
That context-switching cost adds up. Research puts the number at 40-60% of total production time lost to pre-production tasks: ideation, research, scripting, and titling. These 23 prompts compress that significantly. But the fragmented workflow still creates drift between what you research, what you script, and what the algorithm rewards.
For the past year I have been building a workflow that connects those layers. All the research, scripting, outlier analysis, and analytics live in one system with memory of what performed. No more translating insights across tabs.
That tool is Tukey AI. It is purpose-built for YouTube creators who want to use AI across the full research-to-publish workflow, not just the scripting step.
A note on why we built Tukey AI
I spent the better part of a year building the kind of prompt library you just read. It worked. The scripts got sharper. The hook quality improved. The titles started generating actual search traffic.
But every video still required the same setup ritual. Open Claude. Paste the research context. Paste the audience description. Paste the competitor titles. Run the prompts. Copy the output to Notion. Open the analytics dashboard. Try to remember what you did differently last time.
The insight that became Tukey was simple: the tool should know your channel. Not through a settings menu you fill out once and forget, but through every video you research, every script you generate, every title you test. The context should be cumulative, not repeated.
We built Tukey because the prompts alone were not the bottleneck. The workflow around them was.
tukey.ai
FAQ
Do Claude prompts for YouTube creators actually improve video performance? The direct output of a Claude prompt is a better first draft, not a better video. The video performance improvement comes from what that better draft enables: tighter hooks reduce early audience drop-off, stronger titles improve click-through rate, and SEO-optimized descriptions improve search ranking over time. The prompts in this guide are built around those specific outcomes, not just generic "write a script" requests.
What is the best Claude prompt for writing a YouTube script? Prompt 11 in this guide, the Full Script Architect, is the most complete starting point. The key is filling in the audience description and the "transformation" fields with real specificity. A script prompt that says "my audience is beginners" produces generic output. A script prompt that says "my audience is 28-35 year old product managers who have tried journaling before and quit within a week" produces something that feels written for a real person.
How do I use Claude prompts for YouTube titles? Prompt 17 (Title Formula Generator) and Prompt 20 (Title Critique and Upgrade) are the most effective starting points. The key principle is that you need to give Claude the counterintuitive finding and the specific outcome before asking for titles. Without those inputs, Claude defaults to descriptive titles rather than click-directed ones. Descriptive titles describe a topic. Click-directed titles make a promise.
How long does it take to run these Claude prompts for a full video? Research prompts 1-5 take 15-20 minutes the first time, less once you have a repeatable system for collecting competitor titles and comments. Hook prompts 6-10 take 10-15 minutes. Script architecture with prompts 11-16 takes 20-30 minutes depending on video length. The full research-to-publish prompt workflow for a single video runs 60-90 minutes, compared to the 3-4 hours most creators spend on the same work without structured prompts.
Can I use these prompts for YouTube Shorts as well as long-form? The hook prompts (6-10) apply directly to Shorts. The script architecture prompts (11-16) need significant compression: a 60-second Short has no room for a mechanism section or proof section. The most transferable Shorts adaptation is Prompt 9 (Data-Led Hook) combined with Prompt 15 (CTA Optimizer). Research and title prompts apply equally to both formats.
What context should I always give Claude before any YouTube prompt? Three things: your niche in one sentence, your target viewer in two sentences (who they are and what outcome they want), and the specific topic of the video. Claude without audience context produces content for everyone, which means content for no one. Adding those three inputs to any prompt, even the ones in this guide, improves output quality measurably.
The tools are here. The prompts are ready. What you script next is up to you.
My name is EJ Zhang, the CEO at Tukey AI , a production workspace built in your voice. It learns your beliefs and creative fingerprint, surfaces pre-trending topics tailored to you, helps you create with originality, predicts performance before you publish, and learns from every result to make smarter recommendations over time.