Your video can be brilliant, but if the thumbnail is weak, almost nobody clicks. For coaches, that gap is brutal: hours of filming and editing, then a quiet view count. The fix is rarely the video. It is the thumbnail.
This prompt builds YouTube thumbnails for coaches the way the high-CTR ones are actually built: one clear subject, big readable text, a single emotion, brand colors that pop on a phone screen. You hand the AI your video idea and it returns the on-thumbnail text, a paste-ready image prompt, and layout rules. By the end of this page you will also understand why it works, so your next thumbnail is sharper than your last.
When to use this
- You filmed a great video and need a thumbnail that earns the click.
- You keep guessing at thumbnail text and want a curiosity-driven line in seconds.
- You want a consistent, on-brand look across every video without hiring a designer.
- You want two A/B concepts to test instead of betting everything on one image.
- You are repurposing a coaching session, a livestream, or a podcast clip into a YouTube upload.
The prompt
Copy this whole block into ChatGPT, Claude, or Gemini:
You are an expert YouTube thumbnail designer and prompt engineer who specializes in high-CTR thumbnails for coaches and creators. Your job is to turn a single video idea into a detailed image-generation prompt I can paste into an AI image tool (DALL-E, Midjourney, or Gemini), plus the exact on-thumbnail text.
Before writing, ask me up to 3 clarifying questions if anything below is unclear. Otherwise, proceed.
CONTEXT
- My coaching niche: {{NICHE}}
- My channel name / brand: {{CHANNEL}}
- This video's topic: {{VIDEO_TOPIC}}
- The promise or hook of the video: {{HOOK}}
- My brand colors: {{BRAND_COLORS}}
- Visual style I want: {{STYLE}}
- Will I (the coach) appear in the thumbnail? {{FACE}}
TASK
Produce, in this exact order:
1. THUMBNAIL TEXT: 2-4 punchy words that create curiosity or stakes. Give me the main line plus 2 alternatives. No full sentences.
2. IMAGE PROMPT: one paste-ready prompt for an AI image generator describing the subject, composition, expression/pose, background, lighting, color treatment, and where the text should sit. Specify a 16:9 frame at 1280x720 with a clear focal point readable at small size.
3. LAYOUT NOTES: 3 quick rules for placing the text so it stays legible on mobile.
CONSTRAINTS
- Design for legibility at 320 pixels wide (the mobile feed size). High contrast, big text, one focal point.
- Use my brand colors and style; do not invent a logo or fake awards.
- No clickbait that the video does not deliver. The thumbnail must match the hook.
- Keep on-image text to 4 words or fewer. Faces should show a clear, genuine emotion that fits the topic.
- No tiny details, no busy backgrounds, no more than 3 colors competing for attention.
After the three sections, give me 2 alternative thumbnail concepts (a different angle or emotion) I could A/B test.
How to customize it
Replace the seven {{VARIABLES}} before you send it:
| Variable | What to put | Example |
|---|---|---|
{{NICHE}} |
Your specific coaching niche | career coaching for women in tech |
{{CHANNEL}} |
Your channel or brand name | The Confident Pivot |
{{VIDEO_TOPIC}} |
What the video is about | how to negotiate a higher salary offer |
{{HOOK}} |
The promise the video delivers | the exact 3 sentences that got my client a 22% raise |
{{BRAND_COLORS}} |
Your palette | deep navy, warm coral, off-white |
{{STYLE}} |
The look you want | clean, modern, bold sans-serif text, photo of me |
{{FACE}} |
Whether you appear | yes, head and shoulders, smiling and confident |
See it in action (full example)
Here is the exact prompt, filled in for a career coach. This is the whole input, nothing hidden:
You are an expert YouTube thumbnail designer and prompt engineer who specializes in high-CTR thumbnails for coaches and creators. Your job is to turn a single video idea into a detailed image-generation prompt I can paste into an AI image tool (DALL-E, Midjourney, or Gemini), plus the exact on-thumbnail text.
Before writing, ask me up to 3 clarifying questions if anything below is unclear. Otherwise, proceed.
CONTEXT
- My coaching niche: career coaching for women in tech
- My channel name / brand: The Confident Pivot
- This video's topic: how to negotiate a higher salary offer
- The promise or hook of the video: the exact 3 sentences that got my client a 22% raise
- My brand colors: deep navy, warm coral, off-white
- Visual style I want: clean, modern, bold sans-serif text, photo of me
- Will I (the coach) appear in the thumbnail? yes, head and shoulders, smiling and confident
TASK
Produce, in this exact order:
1. THUMBNAIL TEXT: 2-4 punchy words that create curiosity or stakes. Give me the main line plus 2 alternatives. No full sentences.
2. IMAGE PROMPT: one paste-ready prompt for an AI image generator describing the subject, composition, expression/pose, background, lighting, color treatment, and where the text should sit. Specify a 16:9 frame at 1280x720 with a clear focal point readable at small size.
3. LAYOUT NOTES: 3 quick rules for placing the text so it stays legible on mobile.
CONSTRAINTS
- Design for legibility at 320 pixels wide (the mobile feed size). High contrast, big text, one focal point.
- Use my brand colors and style; do not invent a logo or fake awards.
- No clickbait that the video does not deliver. The thumbnail must match the hook.
- Keep on-image text to 4 words or fewer. Faces should show a clear, genuine emotion that fits the topic.
- No tiny details, no busy backgrounds, no more than 3 colors competing for attention.
After the three sections, give me 2 alternative thumbnail concepts (a different angle or emotion) I could A/B test.
And here is the complete output the model returns:
1. THUMBNAIL TEXT
Main line: “+22% IN 3 LINES”
Alternatives: – “SAY THIS, GET PAID” – “THE RAISE SCRIPT”
2. IMAGE PROMPT (paste into your image tool)
“A 16:9 YouTube thumbnail, 1280×720, designed to be readable at 320px wide. Left two-thirds: a confident professional woman in her 30s, head and shoulders, smiling with a knowing, self-assured expression, looking slightly off-camera, wearing a smart coral blazer. Bright, even studio lighting on her face. Background is a clean solid deep-navy panel with a subtle soft gradient, no clutter. Right third reserved for text: large bold off-white sans-serif reading ‘+22%’ stacked over ‘IN 3 LINES’ in warm coral, with a thin off-white outline so it stays legible against the navy. One single focal point (her face), high contrast, three colors only: deep navy, warm coral, off-white. Modern, clean, premium coaching brand feel. No logos, no extra graphics, no busy detail.”
3. LAYOUT NOTES
- Put the text in the right third where her body does not overlap it, and size the biggest word to roughly one-third of the thumbnail height.
- Use one accent color (coral) for the number and white for the words so the eye lands on “+22%” first.
- Add a thin outline or subtle shadow behind the text so it survives both light and dark backgrounds in the feed.
2 alternative concepts to A/B test:
- Reaction angle: Same woman, but mid-laugh holding a phone showing an offer email, expression of pleasant shock. Text: “SHE NEGOTIATED THIS”. Tests curiosity over the proof number.
- Object angle: No face. A clean overhead of a sticky note with three blurred lines and one line in sharp focus, on a navy desk. Text: “THE 3 SENTENCES”. Tests intrigue and works if you would rather not be on camera.
That is ready to generate. Drop the image prompt into your tool, set the text in Canva over the result if you want crisper type, and you have a thumbnail in minutes.
Why this works
Four LLM principles are doing the heavy lifting here. Learn them and every prompt you write gets better:
- Role priming. The opening line (“You are an expert YouTube thumbnail designer and prompt engineer”) tells the model which slice of its knowledge to use. A bare “make me a thumbnail” pulls from the bland average of the internet. Naming a specialist role pulls from the patterns that actually drive clicks: one subject, big text, single emotion.
- Specificity in, specificity out. An image prompt is only as good as the detail you feed it. Vague input (“a nice career video thumbnail”) gives a vague, generic image. Concrete input (coral blazer, navy panel, text in the right third, readable at 320px) gives a usable, on-brand result. The quality ceiling is set by your
{{HOOK}}and{{BRAND_COLORS}}. - Constraints are quality control. The rules about 320px legibility, four words maximum, three colors, and “no clickbait the video does not deliver” each remove a common failure mode. Telling the model what NOT to do is as powerful as telling it what to do, and it is what keeps the output looking designed instead of cluttered.
- Clarifying questions close the gap. The “ask me up to 3 clarifying questions first” line lets the model fill missing detail by asking instead of guessing, which is the single biggest fix for generic AI output. If you forgot to say whether your face appears, it asks rather than inventing a stock photo.
Do this now
- Copy the prompt above into ChatGPT, Claude, or Gemini.
- Replace the seven variables with your real niche, channel, topic, hook, colors, style, and face choice.
- Send it. If it asks clarifying questions, answer them honestly.
- Paste the IMAGE PROMPT into your image tool, then set the THUMBNAIL TEXT crisply (Canva or your editor) and upload. Test the main concept against one alternative.
Pro tips
- Emotion beats explanation. A clear facial expression that matches the hook out-clicks a clever layout. Tell the model the exact feeling: “pleasant shock,” “calm confidence,” not just “happy.”
- Design for the smallest size first. If the text is unreadable at 320px wide on your phone, it loses in the feed. Squint at it before you upload.
- Reuse your winners. Once a color and layout combo earns clicks, lock those variables and only change the text and expression per video. Consistency trains your audience to recognize you.
- Generate the text and the image separately. Let AI design the layout, but set the final type in Canva so it stays razor sharp. Image generators often render text imperfectly.
0 comments
No comments yet.