How to Write With AI Without Losing Your Voice

How to Write With AI Without Losing Your Voice

Open any five AI-written emails, posts, or cover letters and you'll notice something unsettling: they all sound like the same person. Competent, polite, mildly enthusiastic — and completely interchangeable. The usual conclusion is that AI "can't write well." That's wrong. The models are perfectly capable of writing in a distinctive voice. The problem is that almost nobody tells them whose voice to use.

Why AI writing defaults to beige

When you ask a model to "write a follow-up email to a client," it does exactly that: it produces the statistical average of every follow-up email it has ever seen. That average is smooth, safe, and anonymous. It's not a failure of the model — it's a failure of the brief. You asked for a genre, not a voice.

Think of it like hiring a ghostwriter and giving them only the topic. You'll get something publishable, but it won't sound like you, because they have no idea what you sound like. AI is the same, except it will never push back and ask for samples. You have to volunteer them.

What "voice" actually is

Voice feels like a mystical quality, but it breaks down into concrete, describable habits:

• Rhythm and sentence length. Do you write in long, winding sentences that build toward a point — or short ones. Like this. Do you mix the two?
• Vocabulary level. Do you say "use" or "leverage"? "Start" or "commence"? Do you swear? Do you use jargon freely or translate everything?
• Structural habits. Do you open with a story, a question, or a blunt claim? Do you use headings, bullet lists, numbered steps? Do you end with a summary or just stop?
• Recurring phrasing. The little tics: "here's the thing," "worth noting," "in practice." Every writer has them.
• What you never do. Maybe you never use exclamation marks. Maybe you never write "I hope this email finds you well." Absences are part of voice too.

Once voice is a list of observable habits instead of a vibe, it becomes something you can hand to a model.

The fix: show, don't describe

Telling a model "write casually but professionally" barely works — those words mean something different to everyone. What works is giving the model 2–3 real samples of your past writing and asking it to extract the patterns first. Emails, essays, old blog posts, Slack rants — anything written by you, in the voice you want.

The model can then produce a voice profile: a short, readable description of your rhythm, vocabulary, tone, and structural habits. That profile is the brief your ghostwriter was missing. Every piece you generate afterward starts from your baseline instead of the internet's average.

Doing it with the Voice Writer

We built a tool for exactly this. The Voice Writer walks you through it in three steps:

1. Paste 2–3 samples of your own writing. Anything goes — work emails, essays, posts. More variety gives a better profile.
2. Read your voice profile. You get a plain-English summary of how you write: sentence rhythm, vocabulary, tone, structure, recurring habits, and the things you avoid. Most people find this part weirdly interesting on its own — it's like hearing a recording of your own voice for the first time.
3. Generate or polish. Ask for something new ("a follow-up email to a client who's gone quiet") and it comes back in your voice. Or paste a draft — including one another AI wrote — and have it rewritten to sound like you.

The whole thing takes about two minutes, and the difference is immediately visible: the output stops reading like "an AI wrote this" and starts reading like "I wrote this on a good day."

A few honest caveats

Voice-matching is a drafting tool, not a replacement for judgment. You should still read everything before it goes out — the model can match your rhythm but not your facts, and it will occasionally produce a sentence you'd never actually say. Treat it like a very fast first-draft writer who has studied your style, not a clone of you.

And if you're using AI assistance in a context where disclosure matters — graded work, journalism, client deliverables with contractual terms — check the expectations first. Our ethics guide on AI and creative work goes deeper on where the lines sit, and the industry playbooks cover what's considered normal in specific fields.

The short version

AI writing sounds generic because it's answering a generic brief. Give it your voice — real samples, analyzed into a real profile — and the generic problem mostly disappears. Try the Voice Writer with a couple of your old emails and see how it feels. And if your drafts already sound a bit too polished, our companion piece on why AI drafts sound generic breaks down the specific tells to watch for.