How to Keep Your Own Voice When Writing With AI

Most AI writing sounds the same because the model is doing exactly what it was built to do: predict the most average, agreeable next word. Left alone, it produces competent, forgettable copy — the text equivalent of stock photography. Your voice isn't something the tool takes away. It's something you put back in, on purpose, every time. Here's how to do that without it eating your whole week.

Why does AI make everything sound generic?

Because "generic" is the safest bet. An AI model is trained on an enormous average of how everyone writes, so its default is the middle of the road: balanced sentences, tidy transitions, a faint corporate-keynote smell. It has no idea that you say "honestly" too much, that you open with a question, or that your customers love when you're blunt about pricing. None of that is in the prompt, so none of it comes out.

And readers notice. In a 2024 Bynder study of 2,000 US and UK consumers, 50% could correctly spot AI-generated copy, and 52% said they become less engaged the moment they suspect a machine wrote it (Bynder). It gets sharper for brands: when people notice AI in marketing, they're far more likely to trust you *less* — 31% trust the brand less versus just 7% more, per a December 2025 survey of 8,000 consumers run by Klaviyo and Datalily (eMarketer). The generic voice isn't just boring. It quietly costs you the thing small businesses sell on most: being a real person.

How do I train AI to sound like me?

You feed it you. The single biggest upgrade is giving the model real samples of how you already talk, instead of asking it to invent a voice from a label like "friendly and professional." That label means nothing to it, and means something different to every reader.

Pull together three to five short things you've genuinely written or said:

• Two or three of your best old captions or posts — the ones that actually sounded like you.
• A transcript of a voice note where you explain what you do to a friend. Your spoken rhythm is usually your truest voice.
• A real email or DM you sent a customer.
• A short review or testimonial in your customers' words, so the AI picks up the language they use.

Then prompt it directly. Something like: *"Here are samples of how I write. Match this rhythm, vocabulary, and level of formality. Short sentences. No corporate phrasing. Don't add words I wouldn't say."* Paste the samples underneath. You'll feel the difference in the first draft.

Two things make this stick. Save it once. Most tools let you store custom instructions or a saved "voice" profile — paste your samples and rules there so you're not re-teaching the AI every Monday. And name your tics. Write down five things that make your writing yours: maybe you start with a question, never use exclamation points, swear mildly, or always end with a concrete next step. Hand the AI that list. Specific beats vague every time.

What does "edit ruthlessly" actually mean?

Treat every AI draft as a rough draft a junior assistant handed you: useful, fast, and never ready to publish. The voice doesn't come from the generation. It comes from what you change. Ruthless editing is a short, repeatable pass, not a rewrite:

1. Cut the first sentence. AI loves a throat-clearing windup ("In the world of small business..."). Delete it. Your real first line is usually the second one.
2. Swap the tell-tale words. Hunt down and replace the AI fingerprints: *unlock, elevate, seamless, game-changer, dive in, robust, leverage.* Use the plain word you'd actually say.
3. Break the rhythm. AI writes in even, same-length sentences. Chop one in half. Start one with "And" or "But." Add a fragment. Real speech is uneven.
4. Put one true detail back in. A specific number, a customer's name, a thing that happened Tuesday. Specifics are what AI can't invent and readers can't fake.
5. Read it aloud. If it sounds like a press release or a horoscope, you're not done. If it sounds like you talking, you are.

This whole pass takes two or three minutes per post once it's a habit. That's the trade: the AI saves you the blank-page hour, and you spend a few minutes putting yourself back in.

Which parts should I write myself, and which can AI handle?

Let the AI do the labor, and keep the judgment for yourself. A simple split:

• AI is good at: turning a messy voice note into a clean first draft, reshaping one update into versions for different platforms, suggesting hooks you can react to, fixing structure and grammar, and beating the blank page.
• You should own: the opinion, the story, the specific result, the joke, the strong take, and anything a customer would quote you on. If a sentence carries your actual point of view, write or rewrite that one yourself.

A useful rule: AI handles *how it's said*; you own *what's being said and why anyone should care*. The more a piece depends on your taste or your experience, the more of it should pass through your hands.

How do I keep my voice consistent across platforms?

Consistency comes from one source, adapted, not five separate brains. Start every week from a single "what's new" update in your own words: a voice note, a few bullet points, whatever's natural. That update is the source of truth for your voice. Everything else is a translation of it, not a fresh invention.

From that one source, let AI reshape the format for each platform — punchier for TikTok, a touch more context on Facebook, tighter for Threads — while the voice stays anchored to your original words. The trap is starting cold on each platform and letting the AI improvise a slightly different personality every time. Same source in, your editing pass on top, and you stay recognizable everywhere.

This is the workflow Laspi is built around: you record one weekly voice note about what's new and add a few of your own photos, and it turns that into a week of posts written for each platform — Instagram, TikTok, Threads, Facebook — starting from your words, not a generic template. You read, tweak the lines that matter, approve, and publish. The voice is yours because the raw material is yours; the tool handles the formatting and the blank-page part.

How do I know if it still sounds like me?

Use a few quick gut checks before you hit publish. Would you actually say this sentence out loud to a customer standing in front of you? Is there one specific detail in here that no AI could have guessed? Did you cut at least three words the model added? If a regular follower read this, would they know it was you without seeing your name?

If you can answer yes to those, AI is doing what it should: saving you time without flattening you. The model writes the average. Your job, the part that can't be automated, is to make it specifically, recognizably yours. Train it on you, edit ruthlessly, and the voice stays where it belongs.