We use cookies to understand how visitors use our site and improve your experience. Cookie Policy.
AI is a poor writer but an excellent editor. Here's the input that makes the difference — and the one-pass test that strips out everything that doesn't sound like you.
AI-generated LinkedIn content has a recognisable signature. The hooks are formulaic. The paragraphs are balanced to a fault. The vocabulary is slightly too polished. The opinions are well-reasoned and entirely inoffensive. Readers are increasingly good at detecting it — and when they do, trust evaporates.
The problem is not AI. The problem is using AI as a writer when it should be used as an editor. Here is the distinction, and how to use it in practice.
When you ask AI to write a LinkedIn post from a text prompt, it produces a competent average. It has seen millions of posts, found the patterns, and produces something that fits within them. The problem is that fitting within patterns is the opposite of standing out — and standing out is the only thing that grows an audience.
When you give AI something to work with — your words, your specific observation, your actual voice — and ask it to structure and refine rather than generate from scratch, the output retains what made the input interesting while gaining clarity and formatting. That is the right division of labour.
The single most effective technique for making AI-assisted content sound like you is to begin with spoken input. Record a voice note — 60 to 90 seconds, unscripted — on the idea you want to share. Speak the way you would to a colleague. Include the aside that seemed irrelevant. Make the blunt point you would normally soften.
That voice note contains your vocabulary, your rhythm, your specific framing of the idea. When AI converts it into a post, it is not generating a voice — it is shaping one that already exists. The output will sound like you because it started as you.
If you are using a text-based AI tool, the difference between a post that sounds generic and one that sounds like you is almost entirely in what you give it. "Write a LinkedIn post about leadership" produces a generic post. "Here is a specific thing that happened in a meeting last Tuesday, here is what I said, here is what it made me think about leadership — turn this into a LinkedIn post that sounds like me" produces something worth posting.
Context is everything. The more specific your input, the more personal the output. AI cannot invent specificity — only you can provide it.
Every person has verbal and written patterns that make their communication recognisable. Some of these are worth preserving in LinkedIn content — they are part of your voice. Others are worth fixing because they reduce clarity. Learn the difference.
Preserve: your characteristic way of framing a problem, phrases you actually use, the slightly blunt conclusion you tend to reach, the things you are willing to say that most people in your field are not.
Fix: excessive hedging, run-on sentences, the filler phrases you use when speaking but that add nothing in writing, the conclusion buried at the end when it should be at the front.
After AI produces a draft, read it out loud. If there is a sentence you would never say in conversation — a phrase that sounds like marketing copy, a qualification you would not actually make — cut it or rewrite it. One pass of "would I actually say this?" removes most of the artificiality from an AI draft.
The goal is not a post that sounds like it was written without AI. The goal is a post that sounds like you had your clearest, most well-edited writing day. AI can produce that, if you give it the right raw material to work with.
Seven hook formulas built for voice — say them naturally into a memo, and the structure takes care of itself. Your commute becomes your content session.
The decision to read or scroll past your post happens in under two seconds. Here's how to write a first line that earns the click every time.
ChatGPT can write LinkedIn posts. The question is whether it will make you post consistently. An honest look at what works, what doesn't, and what does.
