ChatGPT and Claude can write LinkedIn posts — but without the right prompt, you get generic content that sounds like nobody in particular. Here's the 5-block structure that changes everything.
In short
→ An effective LinkedIn prompt has 5 blocks: role, context, style, constraints, example. → Most people skip the style block and the example — and end up with interchangeable output. → AI has no access to your expertise, tone, or past results unless you provide them. → Without format constraints, AI generates posts that are too long or poorly structured for LinkedIn. → LinkPost automatically calibrates on 27 personalized factors — no need to rewrite your setup every session.
Why generic prompts produce generic posts
"Write me a LinkedIn post about remote management." That's what most people send to their AI. The result looks like the 10,000 other posts written with the same prompt — same bland tone, same 3-bullet structure, same nothing.
LinkedIn rewards engagement. Engagement comes when your audience recognizes YOUR voice — not the voice of an unconfigured AI.
The problem isn't the AI. It's the absence of instructions.
The 5 blocks of a perfect prompt
A solid LinkedIn prompt breaks down into 5 blocks. None of them is optional if you want a draft you can actually publish.
| Block | What it contains | Concrete example | |---|---|---| | Role | Your expertise, sector, background | "You are a B2B management consultant with 10 years on the ground" | | Context | The precise topic and angle | "Post about remote work: what I learned managing 12 people remotely for 2 years" | | Style | Your voice, tics, rules | "Short sentences, no bullet lists, always a number in the hook, no emojis" | | Constraints | Format, length, what to avoid | "Max 1,200 characters, no question at the end, no 'In conclusion'" | | Example | A real post you wrote | Paste a post that performed well for you |
That last block — the example — is the most underused. It's also the one that gives AI the best calibration of your voice. Without it, AI invents a default style that isn't yours.
The mistakes that kill your results
→ Giving the topic without the angle: "management" says nothing. "What I got wrong in my first hire" says everything. → Forgetting format constraints: without instruction, AI generates 2,000 characters of dense prose. → Skipping the style block: without a reference, AI falls back on the interchangeable "professional LinkedIn" tone. → Never giving an example: there's a real difference between saying "write like me" and showing how you write. → Asking for too many things at once: one post, one angle, one intent. Nothing more.
To see what AI actually produces without calibration, read Writing a LinkedIn post with ChatGPT: what you actually get.
The reusable template
Here's the structure to copy and fill out once:
You are [role + expertise + sector].
Write a LinkedIn post about [precise topic + angle].
Intent: [inform / convince / share a lived experience].
Style: [your writing rules: sentence length, tics, go-to formulas].
Format: [max length, desired structure, number of line breaks].
Avoid: [your banned tics, hollow phrases, structures to skip].
Example of a post that sounds like me:
[paste one of your real posts]
Generate the post.
This template takes 5 to 10 minutes to fill out the first time. After that, you keep the style block and the example outside the prompt and reuse them every session.
Why the prompt alone isn't enough
A good prompt produces a good draft. But it doesn't predict whether the post will actually perform. On LinkedIn, the difference between 18 likes and 200 isn't just about writing quality — it's about the hook, the format, the stacked writing tactics, the timing. Our study of 438,413 posts shows that top 1% posts combine 4 to 6 tactics, and 80% of them have a worked hook in the first 200 characters.
Even a perfect prompt can't integrate all those parameters by hand.
The LinkPost AI post generator doesn't just execute a prompt. It calibrates on 27 personalized factors tied to your account, your past posts, and your virality score. You configure it once — then LinkPost applies it to every generation automatically.
If you prefer working from ChatGPT or Claude, you can also connect LinkPost via MCP and generate, predict, and schedule your posts directly from the AI. The setup is explained here: Writing LinkedIn posts from ChatGPT or Claude (MCP).
What to do now
→ Write your style block once: voice, constraints, tics to avoid. → Pull your best-performing post from the last 3 months as your reference example. → Use the 5-block structure every AI session. → Test, measure, and iterate on the style block based on results.
Want to go further? The free post analyzer breaks down what works in your past posts — that's your starting point for calibrating a prompt that actually sounds like you.
These recommendations come from our LinkedIn prompt engineering experience and the data from the internal LinkPost playbook. The quality of the output depends directly on the quality of the instructions you give the AI.
About the author

Yannis Haismann
Co-founder of LinkPost
Yannis writes about LinkedIn content creation, virality prediction and the algorithm. He builds LinkPost, calibrated on more than a million analyzed posts.
See the algorithm studyFree to start · predict virality before you publish