How to Make AI Write in Your Voice, Not Its Own!

Three of my podcast listeners emailed me within two days of each other.

Not to say they loved the newsletter. To ask if I was alright.

The newsletter sounded different. Not bad, exactly. Just generic. The kind of writing that could have been published by a hundred different people in the creator space. None of the specific observations. None of the directness. Nothing that sounded like the person they had been listening to for years.

I had been running my newsletter drafts through Claude for three weeks. I had a prompt that worked well structurally. The sections were right. The length was right. The format was right.

But it did not sound like me. And three listeners noticed in the first 10 days.

Why AI Defaults to Generic

Claude does not have an agenda. It has patterns.

When you give it a topic and a format without specifying how you write, it produces content that matches the most common version of that content. The most common version of a creator newsletter is competent, balanced, informative, and completely interchangeable.

It is not bad writing. It is average writing. And average writing does not build an audience over years.

Your voice is what made your audience choose you over the other 20 creators writing about the same topics. When the AI strips that out, you are left with content that informs without connecting.

The fix is not more editing. More editing just polishes something that is structurally wrong.

What Voice Actually Is

Most creators describe their voice in vague terms. Conversational. Direct. Relatable.

Every creator says that. None of those words help Claude write like you specifically.

Voice is made of specific, observable patterns. The length of your sentences in high-energy sections. The words you overuse when you are making a point you care about. The types of examples you reach for. How you handle counterarguments. Whether you use second person constantly or save it for direct instructions. Whether you quote numbers or avoid them. Whether you occasionally use a blunt closing line or always soften the end.

My voice has specific markers. Short declarative sentences when the point is serious. First-person story opening for almost every piece. Specific numbers over approximations. I never use the word “journey” or “tribe” or “authentic.” I do not hedge with “some people might argue.” I make the argument.

None of that is in a Claude session by default. All of it can be.

Building Your Voice Brief

Spend 30 minutes writing down what you actually do when you write, not what you aspire to do.

Read five of your best pieces. The ones where the response was strongest. Write down what you notice.

How long are your paragraphs in those pieces? How long are the sections? What do you say in the first three sentences? What kind of opening does not appear in any of those five pieces?

Look for specific language patterns. Words you reach for. Phrases that appear in multiple pieces. The way you handle transitions. The way you close.

Write it as a series of rules, not descriptions. Not “conversational” but “second-person address only in instructional sections, not in story sections.” Not “direct” but “no hedging language. State the claim. Do not follow it with qualifications.”

A voice brief of 10 to 15 specific rules is more valuable than a complete prompt with vague tone instructions.

Where Voice Rules Go

Voice rules belong in every prompt that produces finished content.

Not as a single sentence at the end. Woven through the instructions. After you specify the audience. Before you specify the format. Then again as a specific section dedicated to writing patterns.

There is a difference between telling Claude what to write and telling Claude how to write it. Most prompts specify what. The voice rules handle how. Both matter. Most creators only do the first.

After I added voice rules to my newsletter prompt, the next draft needed 20 percent less editing. After I had run it five times and refined the rules based on what was still slipping through, the drafts needed almost no editing.

Three months after those listener emails, one of the same three listeners replied to a newsletter to say it was the best issue I had published all year.

Same topics. Same format. Different voice instructions.

The Rules That Changed the Most

Two rules had the most impact.

The first was a sentence-length rule. I specified that sections building an argument should use short, declarative sentences (subject-verb-object, under 12 words). Explanatory sections could run longer. This matched how I actually wrote and how I had been editing Claude’s drafts every time.

The second was a list of 15 banned phrases that appeared in almost every Claude draft but never in my own writing. Words like “moreover,” “it is worth noting,” “in today’s world,” and “as we can see.” Not banned everywhere in the language. Banned in content that was supposed to sound like me.

Both of those rules eliminated the editing I had been doing most consistently. Which meant they were describing the gap between how Claude defaulted and how I actually wrote.

The gap is not inevitable. You just have to make it visible.

How to Start

Take your last published piece and your last Claude draft side by side.

Find every sentence in the Claude draft that you rewrote during editing. Write down what was wrong with it. Not structurally wrong. Voice wrong.

Those edits are your voice rules. You made them once by hand. Make them permanent by writing them into the prompt.

The Creator’s Super Prompt Library includes a tested voice-rule template that walks through seven specific dimensions of writing voice: sentence structure, person and address, example types, banned phrases, hedging rules, transition patterns, and opening/closing conventions.

Get it here!

Your audience followed you for a reason. That reason belongs in every prompt.