How to find your tone of voice on LinkedIn

Your tone of voice on LinkedIn is two mechanical choices and one hard one. The mechanics are person and address: write from your own experience, and address a reader somewhere in the post. Both show up clearly in our analysis of 12,988 posts. The hard part is specificity, which is the thing that actually distinguishes your writing from everyone else's in the same job: particular incidents, real numbers, opinions you would defend, and the words you use in conversation but delete before publishing.

What is tone of voice on LinkedIn, and what is it not?

Tone of voice is which of your real registers you write in. Everyone has several. The one you use in a board update is not the one you use explaining the same decision to a colleague over coffee, and neither is the one you use texting a friend about how the meeting went. All three are you. Choosing between them is the whole exercise.

What it is not is a brand attribute. Deciding that your voice is “warm, direct and insightful” produces nothing, because those three words describe how you would like to be perceived rather than how you construct a sentence. Nobody has ever improved a post by rereading a values document.

The practically useful definition is narrower: your voice is the set of choices a reader could use to identify you if your name were removed. Which words you reach for. What you are willing to say. What you find funny. Which arguments you have had so many times that you now open with the conclusion. Those are all observable, and all of them show up in writing before any of them show up in a brand guideline.

What do 12,988 posts say about the voice that performs?

Two patterns are large enough to be worth building on, and one popular belief does not survive the data.

First person separates the top from the bottom.Openers that begin from the writer's own experience, using I, my or we, appear in 19.6% of top-decile posts and 10.1% of the bottom half. Roughly double. This is the single most reliable voice signal in the whole dataset.

Addressing the reader separates them too.The word “you” appears somewhere in the visible hook of 47.5% of top-decile posts and 32.7% of bottom-half posts. Note the wording: somewhere in the hook. Actually opening with “you” is rare across the whole cohort and barely differs between groups, 1.6% in the top decile against 1.3% in the bottom half. So the winning pattern is not an instruction at the start. It is a post that begins in the writer's world and turns to the reader before the fold.

The thing that does not matter. Hook length does not separate the groups at all. The median is 206 characters in the top decile and 205 in the bottom half. Neither do numbers in the opening, at 32.1% against 32.8%. If you have been told that a distinctive voice means short punchy lines or leading with a statistic, the data has nothing to say for either.

One more marker worth naming because it is so lopsided: emoji appear in 24.3% of top-decile hooks and 2.7% of bottom-half hooks. That is a nine-fold difference, and it is heavily entangled with format and with who is writing, so treat it as a signal about what kind of accounts do well rather than as an instruction. We look at it properly in what 12,988 posts say about hooks.

Does the right voice change depending on your role?

Yes, and the clearest case is founders. When we compare founder and CEO posts against their own weakest posts rather than against the whole cohort, first-person openers appear in 19.6% of that group's top decile (n=561 posts from 17 authors) and 5.2% of its own bottom half (n=2,810 posts from 12 authors). Nearly four times. Their weakest posts also lean heavily on images, 44.7% against 32.8%, and on hashtag blocks, with four or more hashtags in 39.4% of the bottom half against 13.2% of the top decile.

Read that as a description of two modes rather than as a formula. The founder posts that work are the ones where a person tells you what happened to them. The ones that do not are the ones where a company account borrowed a person's name. Role-specific breakdowns live at our founder and CEO data page.

Does the right voice change depending on your subject?

More than most advice admits, and the within-topic cuts are the interesting ones because they hold the subject constant instead of comparing one topic against everything else.

The finance row is the most striking. In a subject where almost nobody writes from personal experience, the posts that do are six times more common at the top than at the bottom. Where the convention is impersonal, breaking it is worth a great deal.

The productivity row runs the other way, which is why it is in the table. Among productivity posts, the weakest ones address the reader more often than the best ones do. That is what advice-giving second person looks like at scale: numbered instructions telling strangers how to run their mornings. When everyone in a subject is already talking at the reader, adding more of it does nothing, and the posts that stand out go back to first person. The mechanical rules are conditional on what everyone else in your subject is already doing.

The practical instruction: look at the twenty most recent posts in your own subject area, count how many open in first person, and then do the less common thing if you can do it honestly.

What is voice actually made of, once the mechanics are handled?

Person and address get you to competent. Three things get you past it, and none of them can be measured in a dataset of hooks.

Specificity.This is most of it. A sentence that could appear in anyone else's post has no voice in it, no matter how well built. “Hiring is hard” has no voice. “We interviewed eleven people for one role last spring and the one we hired was the only candidate who asked what our churn rate was” is nothing but voice, because nobody else has that sentence. Specificity is not a stylistic preference. It is the only reliable evidence that you were there.

Opinions.An opinion is a claim a reasonable person in your field could disagree with. If nobody could disagree with anything in your post, you did not say anything, and the reader has correctly worked that out by line three. This is the hardest one for people writing under a company's shadow, and it is the difference between being read and being scrolled past. It is also worth being careful with, which is covered in how to write contrarian LinkedIn posts.

The words you would never say aloud. Everyone has a private vocabulary for their own work: the shorthand, the mild exasperation, the joke you make about the third round of feedback. Most of it gets deleted on the way to the publish button and replaced with a professional synonym. That deletion is where voice dies. The word you were about to remove is usually the only word in the sentence that was yours.

What does specificity look like line by line?

The comparisons below are invented for illustration rather than taken from our dataset. Each pair says roughly the same thing. Only one of them could have been written by a particular person.

Notice what is not different between the columns. The right-hand versions are not funnier, shorter, punchier or more confident. They are more particular. That is the entire mechanism, and it is why voice cannot be added at the editing stage. You either wrote down the thing that happened or you did not.

How do you find your own tone of voice?

Four exercises, in the order that produces results fastest.

1. Record yourself for two minutes. Open a voice memo and explain something from your work to an imaginary colleague. Transcribe it. Highlight every word you would not have typed. That list is your voice, and you have been editing it out of your posts for years.
2. Read your own sent folder. Find three emails where you explained something clearly to a client or a colleague. That register, the one where you are trying to be understood rather than to be impressive, is almost always the right one for LinkedIn.
3. Write the post you would not publish. Draft the version with the real numbers, the actual frustration, the thing you think but do not say. Then decide what to remove. Editing down from too honest produces a better post than editing up from safe, every time.
4. Collect your own sentences. Keep a note of lines from your writing that sounded like you. After a month you will see the pattern: your sentence lengths, the rhythm, whether you use questions, whether you are funny in a dry way or not funny at all. That note is a more useful voice guide than anything you could specify in advance.

What none of these involve is choosing adjectives. You are not deciding to be approachable. You are noticing what you already sound like when you are not performing, and then not deleting it.

How formal should your LinkedIn tone of voice be?

Roughly as formal as you are in a good client meeting. Slightly less than an email to a regulator. Considerably less than your company's marketing.

The failure mode is corporate register on a personal profile, and it fails for a structural reason rather than an aesthetic one. Corporate language exists to avoid commitment: it hedges, it uses abstractions so nothing can be pinned to anyone, and it removes the writer. Every one of those properties is the opposite of what makes a personal post worth stopping for. A press release in a feed of people talking is invisible, not offensive.

The other failure mode is worth naming too, because the correction usually overshoots. Deliberate informality, one-sentence paragraphs, faux vulnerability and a story about your Uber driver is also a register, and by now it is a more recognisable one than corporate speak. Both are costumes. The target is neither.

The single test that resolves almost every case: would you say this sentence out loud to someone whose opinion you care about? Not shout it from a stage. Say it, at a normal volume, in a room. If not, rewrite it until you would.

What if your industry will not let you have a voice?

Some constraints are real. If you work in a regulated field, there are things you cannot say about clients, positions, performance or pending matters, and no writing advice overrides that. But most people who believe their industry forbids a personal voice are describing a habit rather than a rule.

The finance numbers in our data make the point. Among finance posts, first-person openers appear in 12.9% of that subject's own top decile (n=279 posts from 24 authors) and 2.1% of its own bottom half (n=1,398 posts from 25 authors). The second-person gap in the same subject is 46.2% against 14.9%. Finance is the most impersonal subject in our dataset by a wide margin, and inside it, the personal posts are the ones at the top. The convention is strong and it is not the thing that works.

Where the constraint is genuine, the usable material moves rather than disappearing:

• Your reasoning, not your clients. You cannot describe the deal. You can describe how you think about that class of decision, which is more useful to a reader anyway.
• Public information, privately understood. Everyone can read the filing. Almost nobody can tell you which line in it matters and why.
• Your own mistakes at a safe distance. Something you got wrong five years ago, at a former employer, with the identifying details removed, is usually clear of every restriction and carries more weight than a current anecdote would.
• What you argue about internally. The disagreements inside your profession are public knowledge to insiders and invisible to everyone else. Taking a side is an opinion that costs you nothing legally.

Check the actual policy before you assume it. The number of people who have never read their employer's social media guidance and are certain it forbids everything is remarkable.

Why do AI-written LinkedIn posts sound flat?

Because a model asked to write about a subject produces something close to the average of everything written about that subject, and the average has no specific incidents and no opinions. It is the exact opposite of the two things that make a voice.

This is not a limitation you fix with adjectives in the prompt. Asking for a conversational tone gets you the average conversational post, which is a recognisable genre with its own clichés. The only real fix is inputs: give the model the actual story, the actual numbers, the thing you said in the meeting. A model can arrange your material. It cannot supply material it was never given, and it will happily generate a plausible substitute if you do not.

There is a useful diagnostic in this for human writing too. If a post you wrote yourself could have been produced by a model given only the topic, then it contains nothing that required you, which is the same problem arriving by a different route. The specific tells readers notice, and what to do about them, are in how to make AI LinkedIn posts sound human, and the detection question is covered in whether LinkedIn detects AI content.

Can a LinkedIn tone of voice be too distinctive?

It can be too performed, which is different from being too strong.

The recognisable failure is a voice assembled from tactics: a provocation in line one, a one-word paragraph for emphasis, a personal revelation that is doing commercial work, and a question at the end that nobody has a real answer to. Each element performs. Together they read as someone doing an impression of a LinkedIn creator, and the audience that notices is the small one that might have hired you.

The test is uncomfortable and reliable. Would a former colleague reading this recognise you, or wince? People who have worked with you are the strictest audience you have and the one most likely to refer you. A voice they do not recognise is a cost, even when the numbers go up.

The other cost is sustainability. A performed voice is exhausting to keep up, and the people who abandon LinkedIn after four months are usually the ones who built a version of themselves they did not want to be. Writing as yourself has the practical advantage of requiring no maintenance.

How do you keep your voice when somebody else writes the draft?

Whether the drafter is a ghostwriter, a colleague or a model, the same three things preserve a voice and nothing else does.

• Supply the raw material yourself. The incident, the numbers, the thing you actually think. Anyone can arrange material. Nobody can invent yours.
• Keep the first and last lines. The opening carries the promise and the close carries the position. If you write nothing else, write those.
• Read every draft aloud before it publishes. Your ear catches wrong register faster than your eye catches anything. Any sentence you stumble over is one you would not have said.

The economics of the ghostwriting version of this are in what a LinkedIn ghostwriter costs, and the structural mechanics of turning a real incident into a post are in LinkedIn storytelling.

How do you check whether a post sounds like you?

1. Read it aloud. Stumbles mark sentences you would not say. Fix those first.
2. Cover your name. Could three other people in your field have written this? If yes, the post has no voice in it yet, and the fix is a specific detail rather than a stronger adjective.
3. Find the disagreeable sentence. There should be at least one claim a reasonable colleague could argue with. If there is not, you have written a summary.
4. Check for one real number or name. A date, a count, a place, a job title. Concrete details are the cheapest proof that you were actually there.
5. Look for the word you deleted. Go back to the draft and find the phrase you softened. Put one of them back.

The caveats apply to every figure above. Our cohort is 12,988 English posts from 65 creators, all with at least 1,000 followers and at least 25 reactions per post, so it describes established accounts. Every relationship is correlational rather than causal: first-person openers are more common among the best posts, which is not the same as writing in first person causing a post to do well. Engagement counts are lifetime-cumulative at the time the data was collected and post ages vary. Most relevant to this article, the text findings describe the visible hook above LinkedIn's fold rather than complete posts, so they measure how people open rather than how they write throughout. Full methodology is in our study of 34,000 LinkedIn posts.

The short version on finding your tone of voice on LinkedIn

Find your tone of voice on LinkedIn by writing from your own experience, addressing a reader somewhere before the fold, and then being specific enough that nobody else could have written the sentence. The mechanics are visible in the data: first person appears in 19.6% of top-decile posts against 10.1% of the bottom half, and the word “you” in 47.5% against 32.7%. The rest is not measurable and matters more. Record how you talk, keep the words you would normally delete, say at least one thing a colleague could argue with, and read every draft aloud before it goes out.