Writing Voice: How to Find Yours (and Keep It When You Use AI)

You've read two posts on the same topic this week.

One you forgot before you finished it. The other you sent to a friend.

Same subject. Same facts, more or less. The difference wasn't the information. It was the voice. One sounded like a press release. The other sounded like a person who had actually thought about the thing and couldn't wait to tell you.

That's the whole game. And it's the one thing a tool can't hand you.

So here's how to find your writing voice, how to name it so you can repeat it on purpose, and how to keep it when AI starts doing the grunt work.

What your writing voice actually is

Your writing voice is not a vibe. It's a set of choices you make so consistently that readers could recognize you with your name removed.

It's the words you reach for. The length of your sentences. The jokes you can't help making. The thing you always circle back to. The opinion you hold a little too strongly.

Most people think voice is something mysterious you either have or don't. It isn't. It's a pattern. And patterns can be found, named, and reused.

The reason it feels mysterious is that you can't see your own pattern from the inside. You're too close. The same way you can't hear your own accent, you can't hear your own voice on the page. You need a way to step outside it.

How to find your writing voice

Voice is not invented. It's excavated. You already have one. The work is getting it out of your head and onto the page without sanding it flat on the way.

Write the way you talk. Open a doc and explain your topic to one specific person, out loud, then type what you said. Not "compose an article." Talk. The cadence you use when a friend asks "wait, why does that matter?" is your voice. The cadence you use when you're trying to sound smart is somebody else's.

Read it aloud. This is the single fastest voice tool there is. The sentences you stumble over are the fake ones. The places you'd never actually say those words to a human face. Cut them. The lines that sound like you when you say them are the keepers.

Name your tics. Go read five things you've written. Look for what repeats. Do you open with a tiny scene? Use a lot of short fragments? Lean on one-line paragraphs for emphasis? Always end on a turn? Those repeated moves are your voice, made visible. Write them down. Now they're a recipe instead of an accident.

Cut the borrowed phrases. "In today's fast-paced world." "At the end of the day." "It's important to note." Nobody talks like this. These are the phrases that slide in when you stop writing and start performing writing. Every one you cut, your real voice gets one notch louder.

There's a question a UK writer named Katie Skelton asks the business owners she advises, and it works on voice too: what would you say to a friend who came to you with this? You'd never answer a friend in corporate-speak. You'd be direct, specific, a little funny. That's your voice. The trick is writing as if the reader is that friend.

Keep one reader in mind, not the algorithm

Reach is down everywhere right now, and the loudest advice is to flatten yourself to please the feed. Post more. Hedge your takes. Sound like everyone who's winning.

That's exactly how you lose your voice.

Here's a more honest frame, also from Skelton: you only need one reader who thinks what you wrote is brilliant. Write for that one person. Be specific enough, opinionated enough, you enough that one human feels seen.

A voice that tries to please everyone reads like it came from nobody. A voice aimed at one real person is the one strangers recognize and follow.

Where AI flattens your voice (and how to stop it)

We build an AI tool, so this is the awkward part to admit. But pretending otherwise would be its own kind of fake.

AI models are trained to be agreeable and average. They predict the most likely next word, and the most likely word is, by definition, the unsurprising one. Your voice lives in the unlikely words. The weird specific detail. The fragment that breaks the rhythm on purpose. The opinion you weren't sure you should say out loud.

Hand the whole job to a model and it smooths all of that away, because smoothing is what it's built to do. You get the average of everything ever written on the subject. Competent. Forgettable. The exact texture readers have learned to skip.

So don't hand it the parts that are supposed to be yours.

• The point of view is yours. AI argues any side with equal conviction because it has no skin in the game. The take has to come from you.
• The specific details are yours. The number, the name, the moment from last Tuesday. AI invents plausible-sounding specifics, and plausible-but-false is the worst thing you can publish.
• The final read-aloud is yours. The places you stumble are the places a model flattened you back to average. Rewrite those in your own words.

Let AI take the busywork: the blank-page rough draft you'll mostly discard, the outline, the reshaping of one piece into the formats you need. Keep your hands on the thinking and the voice. Mechanics get automated. Judgment stays human.

Make your voice survive the reformat

Here's the trap that quietly erases voice: you write something good, in your real voice, then you rewrite it from scratch for your newsletter, then again for an X thread, then again for LinkedIn. Four rewrites, four chances to flatten yourself back to average, usually late at night when you're tired and willing to settle.

The fix is to write once, in your voice, and reshape the same words for each room instead of rewriting them.

That reshaping is where Writeous lives. Paste one markdown file, the one you wrote in your own voice, and get back a blog post, a newsletter, an X thread, and a LinkedIn post, each formatted right for its platform, in about a minute. It reshapes the words you already chose. It doesn't reach in and average out your voice. Free to try, no login.

We'll name the limit too: when your blog runs on Ghost, Writeous re-syncs the published post in place, so editing your source updates what's live. Social publishing through Typefully works, but a sent post can't be edited after it's out, so that's best-effort, not true sync.

The takeaway

Your writing voice is a pattern, not a gift. Find it by talking instead of performing, reading aloud, and naming the moves you already make. Protect it by writing for one real reader, not the feed.

And when you bring in AI, keep the point of view, the real details, and the final read for yourself. Hand it the grunt work, never the voice.

Do that, and your writing keeps sounding like a person. Because it is one.

Write once. Publish everywhere. Actually.