AI and LinkedIn ·
LinkedIn prompts: when they help and when they dilute you
There are thousands of threads promising "the definitive prompt for growing your LinkedIn". Copying and pasting those prompts produces homogeneous, recognisable posts reproduced by hundreds of profiles at once. What looks like a shortcut turns into mass dilution.
This guide is for professionals who do want to use AI in their personal branding work without losing their own voice. We are going to separate the prompts that help you think better (which are worth it) from the prompts that write for you (which erode a personal brand in the medium term).
By Sheena de PunkVoice · Edited by Mario Pérez
Two categories of prompt: thinking vs writing
The prompts useful for LinkedIn fall into one of two very different categories. The ones that help you think better before writing (they question your idea, give you counterarguments, suggest angles that had not occurred to you). And the ones that write for you (they generate the whole post, reword your draft, produce viral variants).
The first increase your editorial judgement. The second replace it. The distinction is not minor: in the medium term, professionals who use AI to think publish better over time, and those who use it to write publish worse over time.
The paradox explains why so many profiles that promised growth with generalist AI have watched their engagement fall within six months. The apparent efficiency consumed the very asset holding up their reach: their own voice.
Three prompts that do help
The questioner. "Read this draft and tell me three things a sceptical reader could object to. Do not rewrite, only question." It forces you to anticipate counterarguments before publishing. The post comes out stronger because you already thought the objections through in private.
The cliché detector. "Read this text and flag every sentence that could have been written by any other professional in my industry. Do not correct, only mark them." It pushes you to rewrite those sentences in your voice, avoiding the generic template tone.
The consistency checker. "Here is today's post and three earlier posts of mine. Are there inconsistencies between what I argued before and what I am arguing now? Where?" It warns you when you are contradicting previous positions without having noticed, something an attentive audience does spot.
All three share a structure: the AI thinks with you, it does not write for you. The useful output is an analysis, not a replaceable text.
Why a ChatGPT post gets spotted (and how not to fall into it)
Three prompts that destroy your voice
The full generator. "Write a LinkedIn post about topic X in a professional but warm tone." The output is generic by definition: a generalist AI produces average text for an average audience. Publishing it puts you on the same plane as hundreds of profiles who used the same prompt.
The rewriter. "Take this draft of mine and improve it." The AI's concept of "improving" is standardising: it corrects your vocabulary, softens your contradictions, makes your tone generic. What it hands back is your post run through the style manual's shredder. What you lose is what made you unique.
The viral generator. "Give me five variants of this post optimised for LinkedIn." The AI applies the known viral markers: an opening with a figure, the "it is not X, it is Y" pivot, a close with a rhetorical question. The post reads as a template within the first few seconds.
All three share another structure: the AI replaces your editorial judgement. The output is a text anybody could sign, and that is exactly the opposite of a personal brand.
Grey areas: using AI but carefully
A very rough first draft when you are blocked. Asking AI for an initial outline when you cannot find a way in can unblock you. Important: use the outline as a starting point, rewrite the whole text in your voice, and do not copy a single sentence literally.
Translating one of your posts into another language. If you publish in Spanish and want an English version, AI translates decently. Important: review it manually, because generalist AI markers in English are even more visible than in Spanish. A literal translation falls short, it is worth adapting the cultural tone too.
Summarising your own old posts for a thread. When you want to condense three earlier posts into a new one, AI can help you identify key points. Important: the final post has to be entirely yours, the AI only helps you remember what you said.
Generating topic ideas. A brainstorm of 20 angles on a topic you know well. Useful for finding angle 18 that had not occurred to you. Important: only select the angles that genuinely resonate with you, do not publish them because the AI generated them.
Why viral prompts do not work twice
The cycle is predictable. Somebody shares "the prompt that changed my LinkedIn" in a viral thread. Thousands of profiles try it the same week. The feed fills with posts with the same structure, the same rhythm, the same recognisable markers. The audience detects the pattern and discards the content en masse. The prompt stops working.
The next viral prompt repeats the cycle. And another, and another. Professionals chasing viral prompts live in a degradation loop: every new prompt pays off less than the last, because the audience trains itself faster at detecting them.
The way out of the loop is stopping the search for the magic prompt. The asymmetry: genuinely original content (even if it takes longer to produce) does not compete with templates, because it operates on a different plane. While everyone else differentiates on which prompt they use, you differentiate on what you think.
The structure of a useful prompt when what you want is to think
A prompt that helps you think usually has three elements: context (what topic, for what audience, what starting point), a specific request (what kind of analysis you want, not what kind of text), and a constraint (what you explicitly do not want the AI to do).
A correct example: "I am writing a LinkedIn post about why engagement pods erode credibility. My audience is mid-market B2B professionals. I have written this draft (...). Give me three specific objections a sceptic of my thesis could raise, without rewriting the text. Do not suggest style improvements, only substantive objections."
An example to avoid: "Improve this post about engagement pods so it gets more engagement on LinkedIn."
The first gives you a usable analysis that enriches your draft. The second hands your draft back with AI style and you lose your own voice.
What changes and what stays the same
The models will keep improving. What is detectably AI text today will have fewer visible markers within a year. Some prompts that produce recognisable output today will soon give more camouflaged results.
What does not change: using AI to think will always pay off more in the medium term than using it to write, whatever the model. This is not a question of the output's technical quality, it is a question of which asset you build on your profile.
If your asset is your editorial judgement, using AI as a companion amplifies it. If your asset is your own voice, delegating it to an AI erases it. That logic holds with ChatGPT, with Claude, with Gemini and with whatever model comes next.