When to use it
Use this when you have something genuinely worth saying, a lesson learned the hard way, a contrarian take, a customer moment, a number that surprised you, but you're staring at the blank box. It earns its keep by turning your half-formed thinking into a postable draft in minutes, while preserving the one thing that makes the post worth reading: your actual point of view. Founders, operators, and sellers building a personal brand or social-selling get the most from it.
Do NOT use it to manufacture an opinion you don't hold, the model will produce a confident, hollow, instantly-forgettable post, and your audience can smell it. If you don't have a real POV or a real moment to anchor it, you don't have a post yet. Also do not post the output verbatim; the final voice pass is non-negotiable because people who follow you know how you actually talk, and AI cadence reads as performance.
The principle: a LinkedIn post lives or dies on one sharp idea and a hook that stops the scroll. The most common failure is sanding a strong, specific opinion into safe, hedged mush to avoid disagreement, but disagreement is distribution. This prompt is built to preserve the edge of your take and force specificity (a real number, a real moment) rather than generic thought-leadership.
The prompt
Prompt, paste into Claude, ChatGPT, Gemini
You are a ghostwriter for sharp B2B operators. You write LinkedIn posts that sound like a smart person talking, not a marketer performing. You despise cringe: no emoji walls, no 'Agree?', no fake humility ('humbled to announce'), no engagement-bait, no broetry where every. word. is. its. own. line for no reason. You never invent details, numbers, or quotes, you work only from what I give you.
CONTEXT
My raw POV / story (rough, bullets fine): {{RAW_POV}}
Who I want to reach: {{AUDIENCE}}
What I want them to think or do after reading: {{GOAL}}
My voice / how I actually talk: {{VOICE_NOTES}}
Format preference (story, contrarian take, lesson, teardown): {{FORMAT}}
TASK
Turn my raw thinking into a LinkedIn post that keeps my point of view intact and sharp.
METHOD
1. Find the ONE idea worth posting. If my input has three ideas, pick the sharpest and cut the rest, a post with one clear point beats a post with three muddy ones.
2. Preserve my actual opinion, including its edge. Do not hedge a strong take into a safe one.
3. Anchor it in a concrete moment, number, or example from MY input. Never invent one.
4. Write a first line that stops the scroll without clickbait or 'In today's fast-paced world.'
5. Build for mobile reading: short lines, white space, one idea per line cluster.
6. End with a genuine discussion prompt or a clean takeaway, not a desperate CTA.
OUTPUT FORMAT
1. The post (120-220 words)
2. Two alternate first-line hooks
3. One comment to seed in the thread to drive discussion (often a supporting number or a counterpoint)
CONSTRAINTS
- 120-220 words. One idea.
- Specific, not generic. Use a real detail from my input or ask me for one before writing.
- Banned: hype words (unlock, leverage, supercharge, seamless, game-changer, revolutionize), 'In today's world', 'Let that sink in', 'Agree?', emoji spam.
- Max 3 hashtags, only if they genuinely aid discovery (usually skip them).
- If my RAW_POV has no concrete detail or no real opinion, say so and ask me one question to sharpen it before writing.Run it from the terminal
$# the raw POV lives in a file so it's easy to edit between runs
$ls prompts/linkedin-pov.md
prompts/linkedin-pov.md
$# shape a raw contrarian take into a post with Claude
$claude -p "$(cat prompts/linkedin-pov.md)"
1. The post:
We deleted a 7-email nurture sequence last quarter.
Replaced it with one good email and a call from an actual human.
Conversions went up.
Here's what I think happened: the sequence wasn't nurturing anyone. It was making
us feel busy. Seven touches of 'just checking in' that nobody opened past email two.
Most nurture isn't broken because it's too simple. It's broken because it's too
automated to be worth reading.
If your sequence has more emails than ideas, that's the problem.
What's the longest sequence you've ever quietly killed?
2. Alternate hooks:
- 'More automation isn't more nurture. Usually it's less.'
- 'Our best-performing nurture email had a sequence length of one.'
3. Comment to seed: Drop the before/after open rates in the first comment.
$# edge preserved; do the final voice pass yourself before posting
$
Variables to swap
| Variable | What to put there |
|---|---|
{{RAW_POV}} | Your opinion, lesson, or story in rough form, bullets are fine |
{{AUDIENCE}} | Who you want to reach (e.g. RevOps leaders, early-stage founders) |
{{GOAL}} | The reaction or action you want from readers |
{{VOICE_NOTES}} | How you actually talk, direct, dry, plainspoken, a little contrarian |
{{FORMAT}} | Story, contrarian take, lesson, or teardown (optional) |
Worked example
Input
Filled-in example
RAW_POV: We killed our 7-touch nurture sequence and replaced it with one good email + a call from a human. Conversions went up. People overcomplicate nurture.
AUDIENCE: Demand gen and marketing ops leaders
GOAL: Make them question their bloated automation
VOICE_NOTES: Direct, a little contrarian, no fluff, dry humor
FORMAT: Contrarian takeOutput
What the model returns
1. The post:
We deleted a 7-email nurture sequence last quarter.
Replaced it with one good email and a call from an actual human.
Conversions went up.
Here's what I think happened: the sequence wasn't nurturing anyone. It was making us feel busy. Seven touches of 'just checking in' that nobody opened past email two.
The one email worked because it said something useful. The call worked because people would rather talk to a person than get drip-fed.
Most nurture isn't broken because it's too simple. It's broken because it's too automated to be worth reading.
If your sequence has more emails than ideas, that's the problem.
What's the longest sequence you've ever quietly killed?
2. Alternate hooks:
- 'More automation isn't more nurture. Usually it's less.'
- 'Our best-performing nurture email had a sequence length of one.'
3. Comment to seed: Drop the before/after open rates in the first comment, a real number turns a take into evidence and gives people something to argue with.Tips to get more out of it
- Feed it your actual numbers and moments. Specificity is what separates a credible post from generic thought-leadership; 'conversions went up' is fine, but '7 emails to 1, conversions up' is a post.
- The first line is 80% of the job. If none of the three hooks grab you, ask for five more and specifically request one that 'states the conclusion before the setup.'
- Always do the final voice pass yourself. The fastest way to lose a personal brand is to post in a cadence your audience knows isn't yours, read it aloud and rewrite the lines that sound like a model.
- Resist the urge to add a second idea. If the post feels thin, sharpen the one idea rather than bolting on another; LinkedIn rewards focus.
- Skip the hashtags. On LinkedIn they rarely move reach and they make the post look like it's trying. The model will offer them; usually cut them.
- Across models: Claude tends to preserve a contrarian edge well; ChatGPT often softens it toward consensus (tell it 'keep the spicy version'); Gemini can over-structure into a listicle (ask for prose).
- Reuse the strongest single line as a hook for an ad or a cold email later, a good POV travels across formats, and you've already pressure-tested it on an audience.