How to Get AI to Write in Your Brand Voice

I run AI workshops for founders and marketing teams, and there’s one complaint I hear more than any other: “Everything AI writes sounds the same.”

They’re right. You ask AI to draft a blog post and it comes back sounding like a corporate handbook written by a committee. You try again — “make it more casual” — and now it reads like a college freshman’s cover letter. So you spend an hour editing, and halfway through you think, I could have just written this myself.

AI doesn’t have bad taste. It has no taste at all. When you don’t tell it who to be, it defaults to “generic helpful assistant.” And that’s nobody’s brand.

The fix is building what a brand voice skill — a reusable document that teaches AI exactly how you write and what you avoid. Build it once, and every draft comes out closer to final. Less editing, more publishing.

I’m going to walk you through the exact process I use with clients. It works whether you’re a solo founder or a company keeping fifty writers on the same page. And it works with any AI tool you’re already using.

Step 1: Feed AI your existing work

If you’ve been writing for any stretch of time, you already have the raw material for your brand voice. It’s sitting in your drafts folder, your sent mail, your blog archives. You just haven’t organized it.

Pull together 5-10 pieces of your best writing. Blog posts, emails, memos, newsletter issues, social posts — whatever. The key is picking the ones where you sound most like yourself. The pieces that made you think “yeah, that’s me.”

Drop them into a conversation with Claude or ChatGPT and use this prompt:

Review these writing samples and analyze my voice. Identify patterns in: sentence length and rhythm, vocabulary choices, level of formality, use of humor, how I open and close pieces, how I handle transitions, and any distinctive quirks or habits. Be specific — give me examples from the text.

AI is genuinely good at spotting patterns in writing. It’ll catch things you didn’t even realize you were doing — like always opening with a question, or ending every section with a one-line punch, or using way more contractions than you thought. It sees the patterns because it doesn’t have the blind spots you do about your own writing.

If you don’t have a body of existing work yet, that’s totally fine. Skip ahead to Step 2. The interview process can build your voice from scratch.

Step 2: Have AI interview you

The analysis from Step 1 gets you maybe 60% of the way there. The other 40% lives in your head — preferences you’ve never written down, rules you follow without thinking, pet peeves you didn’t know you had until someone violated them.

This is where you tell AI to flip the script and interview you:

Based on the voice analysis, I want to build a complete brand voice document. Interview me to fill in the gaps. Ask me one question at a time about my writing preferences, my audience, words or phrases I love or hate, and anything else you need to capture my voice accurately.

Let the conversation run. It’ll ask things like: Do you use em dashes? How do you feel about exclamation points? Do you lean into jargon or avoid it? Sentence fragments for emphasis — yes or no?

The trick is being specific with your answers. “I don’t like formal language” is fine as a starting point. But “I use contractions, I skip semicolons, and I write like I’m explaining something to a smart friend over coffee” — that’s the kind of input that builds a voice skill with real teeth.

Step 3: Generate the voice skill

Now you ask AI to turn the analysis and the interview into one structured document:

Based on the analysis and our conversation, create a brand voice skill document. Organize it into these sections: Tone, Sentence Rhythm, Authentic Voice, Structure, Formatting, Emotional Tone, and What to Avoid. For each section, include specific rules and examples from my actual writing. Make every rule concrete and actionable.

A simplified version of the output looks something like this:

# Brand Voice — [Your Name]

## Tone
- Conversational and approachable -- peer-to-peer, not authoritative
- Like learning from a friend who's figured some stuff out
- Not sterile or overly formal

## Sentence Rhythm
- Mix short punchy sentences with longer explanatory ones
- Short hits home: "That's it. You're done."
- Long unpacks the idea and gives it room to breathe

## Formatting
- Use contractions (don't, won't, it's)
- Em dashes sparingly -- save them for impact
- No exclamation points in body copy

## What to Avoid
- Corporate buzzwords (leverage, synergy, utilize)
- Opening with "In today's [anything]..."
- Passive voice unless it reads more naturally
- Making the reader feel behind or overwhelmed

Save this somewhere accessible. In Claude, add it to your project instructions so it applies to every conversation automatically. In ChatGPT, paste it into custom instructions or memory. Load it once, use it everywhere.

Step 4: Test and refine

Your voice skill won’t be perfect on the first try. Mine wasn’t either.

Run 3-5 test outputs. Ask AI to write a short email, a blog intro, a social post — all using your voice skill. Then put the results next to your actual writing and read them side by side.

You’re looking for the misses. Maybe it’s reaching for words you’d never use. Maybe the rhythm is too choppy, or the tone is right but the structure feels off. These gaps are normal; they’re how the skill gets good.

Each time you spot one, add a specific rule. And I mean specific. Don’t write “be more natural.” That kind of vague direction produces vague writing. Instead: “never use the word ‘utilize’ — say ‘use’” or “always open blog posts with a personal anecdote, not a statistic.”

Vague instructions produce writing that sounds like anybody. Specific instructions produce writing that sounds like you.

Step 5: Chain skills for better output (advanced)

Once your voice skill is solid, you can push it further by chaining it with other writing tools. This is the approach I use for my own content.

I run three passes on everything I write with AI:

First, a direct draft. Core content, focused on clarity and value. I don’t think about voice at all — just get the ideas down and the structure right.

Second, a humanizer pass. I run the draft through a humanizer skill that catches common AI writing patterns and strips them out. The filler transitions, the “it’s important to note” hedging, that AI paragraph structure where every point gets identical treatment.

Third, a brand voice pass. I apply my voice skill to the cleaned-up draft. This is where it goes from “AI-generated” to “sounds like I actually wrote this.”

Each pass builds on the last. The direct draft captures what I want to say. The humanizer strips the AI residue. The voice pass makes it mine. No joke, this workflow has made me at least 3 times faster at producing content I’m actually proud to publish.

Worth the extra steps? A 2024 HubSpot report found that 86% of marketers edit AI content before publishing. Most of that editing is the same thing over and over — making it sound less robotic. A multi-pass workflow does that editing automatically.

Why this matters

Readers can tell when something feels off. They might not name it, but they feel it. A 2024 Bynder study found that 52% of consumers disengage when they suspect content is AI-generated. Another 26% called AI copy “impersonal.”

Your brand voice is what separates your content from the flood of AI-generated sameness. And building a voice skill is a one-time investment. You do the work once — a couple hours, tops — and every piece of content gets better from that point forward.

FAQ

Do I need a lot of existing writing? Nope. Five to ten pieces is plenty to get started. If you don’t have anything yet, the interview in Step 2 builds your voice from scratch. You’ll just need a couple extra rounds of refinement.

Does this work for a company, not just a person? Same process, just with bigger inputs. Gather writing samples that represent how the company should sound, interview the people who own the voice, and build the skill the same way.

Which AI tool should I use? Whichever one you’re already using. Claude and ChatGPT both support custom instructions where you can load your voice skill once. The method is tool-agnostic.

How often should I update it? Update whenever you spot a pattern the skill doesn’t capture. Most people make 5-10 tweaks in the first week, then occasional updates as their voice evolves.