Make AI-Written LinkedIn Posts Sound Like You

Most people can spot an AI-written LinkedIn post within two lines. Not because the writing is bad — it is usually clean and well organised — but because it is the same. The same shape, the same rhythm, the same three-word sentence used as a paragraph for emphasis. Once a reader has seen that pattern fifty times in a week, the fifty-first stops landing, no matter how good the point is.

The usual reaction is to give up on AI drafting altogether. That is an overcorrection. The sameness comes from a fixable cause: a generic prompt gets a generic answer. If the model is told nothing about how you write, it falls back on the average of everything it has read, and the average of LinkedIn is exactly the voice everyone is tired of.

Why do AI-written LinkedIn posts all sound the same?

Ask any assistant to “write a LinkedIn post about X” and you will get a recognisable set of habits, because that is what the request invites:

• A hook that announces itself. A short, dramatic opening line on its own, then a line break, then the real start of the post.
• A rhetorical question in the first three lines. Often one nobody was actually asking.
• Tidy parallel structure. Three bullets, each the same length, each starting with the same part of speech.
• An em dash habit and a “this is not X, it is Y” sentence. Both are fine once. Both are a tell when they appear in every post.
• A closing question asking for comments, then five hashtags whether or not you use hashtags.

None of that is wrong in isolation. The problem is that it is a template, and templates are visible in a feed where everyone is using the same one. Your own writing has features too — you probably start with a plain statement of fact, or run longer sentences, or never use hashtags at all — and those features are what make a post read as yours. The model simply does not know them unless you say so.

What is a voice profile, and how does it work?

A voice profile is a short description of how you want to sound, saved once and applied to every future draft. In Kalovio it is two tools: set_voice_profile to save it and get_voice_profile to read back what is currently stored. You write it in plain language, in a normal chat message. There is no form to fill in.

The point is that it is saved. Pasting style instructions at the top of a prompt works, but only for that one message, and only if you remember. A stored profile survives the conversation. Open a new chat next week, ask for a post, and the same rules apply without you typing them again. That is the difference between an AI writing assistant and a LinkedIn post generator that actually sounds like you.

Useful things to put in it:

How do I describe my own writing voice properly?

Most people write a voice profile that is too polite to be useful. “Professional but approachable” describes roughly everyone on LinkedIn, so it changes nothing. Three habits make a profile actually bite.

Be specific enough to be falsifiable. “Conversational” is vague. “Short sentences, no more than twenty words, and I never use a rhetorical question as an opener” is a rule the model can follow and you can check. If you cannot look at a draft and say whether the rule was obeyed, the rule was too soft.

Say what you do not do. Negative instructions carry a lot of weight because they remove the defaults. “No emoji. No hashtags. Never open with a one-line hook followed by a blank line. Do not end by asking for comments.” That alone strips out most of the recognisable AI shape.

Use your own posts as evidence, not as vibes. Open three posts you were happy with and look for what they share. How long is the first sentence? Do you use bullets or prose? Do you name people or keep it abstract? Do you tell a small story before the point, or state the point first? Write down what you find, not what you think you sound like — the two are usually different.

A profile that does this ends up looking roughly like: direct and slightly dry, no emoji, no hashtags, one to three short paragraphs, always open with a concrete fact or a number, British spelling, topics are supply-chain software and hiring, never use the phrases “game changer” or “let that sink in”. That is enough to change every draft you get.

Can it learn my voice from my old LinkedIn posts?

Partly, and this is where being precise matters more than being impressive.

Kalovio has a tool called learn_voice_from_history. It reads posts you have already published and works out patterns — typical length, whether you use emoji, how you tend to open — and turns them into a voice profile you can then edit.

The practical consequence is that the feature gets better the longer you use it. Write your profile by hand at the start, because that is the part that works on day one. After a month or two of posting through Kalovio, run learn_voice_from_history and see whether it noticed anything you did not think to write down. Treat its output as a draft profile: read it, correct the bits it got wrong, and save it with set_voice_profile.

It is worth saying plainly that any tool claiming to read and imitate your full LinkedIn back catalogue is either scraping your account or overstating what it does. The API route to that data is closed.

How long can a LinkedIn post be?

The limit is 3000 characters — characters, not words, so spaces, line breaks and emoji all count. That is the ceiling LinkedIn enforces, and it applies whether you write the post yourself or generate it.

Two things follow. First, put your typical length in your voice profile, because “as long as it needs to be” reliably produces something longer than you wanted. Second, if a draft comes back near the limit, that is usually a sign the model padded rather than that the idea needed the room. Ask for it at half the length and see what survives — the version that survives is normally the better post.

LinkedIn also collapses posts after the first few lines behind a “see more” link, so the opening lines carry disproportionate weight. That is another reason to tell the profile how you open, rather than leaving it to the default dramatic hook.

Should I publish the first draft?

No, and a good voice profile does not change that. What it changes is the size of the edit — from rewriting the whole thing to fixing two sentences.

A workflow that holds up:

• Ask for the post, then read it out loud. The lines you stumble on are the lines that are not yours. That test is faster and more reliable than staring at the screen.
• Fix by instruction, not by regeneration. “Cut the opening line and start with the second paragraph” gets you a better result than “try again”, which usually returns a different post with the same problems.
• Feed corrections back into the profile. If you delete the same thing three times — the closing question, the emoji, the word “leverage” — that is not an editing job, it is a missing line in your voice profile. Update it with set_voice_profile and stop fixing it by hand.
• Park it as a draft when you are unsure. create_draft stores it so you can come back with fresh eyes rather than publishing to get it off your desk.
• Approve the exact wording before it goes out. With Kalovio this is not optional. Publishing, scheduling and queueing are two-step: the first call shows you a preview with a signed token covering the exact text, and nothing is sent until you confirm.

That last step doubles as a quality gate. Because you have to look at the final wording before anything publishes, the “post it and hope” failure mode is not available.

What this actually buys you

A voice profile does not make AI write like you. It makes AI stop writing like the average of LinkedIn, which is most of the distance. You still supply the idea, the specific example and the judgement about whether the post is worth publishing. What you stop supplying is the same paragraph of style instructions, typed again, every single time.