How to Make Claude Write in Your Voice (Without the AI Slop)

You've been using Claude for a few weeks and something still feels off. The content comes out technically correct, reads fine, checks all the boxes, and sounds absolutely nothing like you. It sounds like a well-behaved LinkedIn post from someone who's never had a strong opinion in their life. That's not AI slop in the obvious way, not the em dashes and the "of course!" openers that scream ChatGPT, but it's still not you. And your audience can feel it.

Making Claude write in your voice is 100% possible, but it doesn't happen by accident and it doesn't happen just because you switched from GPT. Claude needs to be trained on who you are, how you think, and what you'd never say. Without that setup, you'll keep getting content that's clean but forgettable, and forgettable doesn't convert.

Here's exactly how to do it.

Why Claude Defaults to Generic (Even When You Ask It Not To)

Claude is trained to be helpful, harmless, and honest. That's great for a lot of things. It's terrible for content that's supposed to polarize, provoke, and make the right people stop scrolling and DM you. Without specific instructions, Claude defaults to a neutral, balanced, slightly formal tone because that's what "helpful" looks like in its world.

You've probably noticed this. You ask it to write a caption and it gives you something that could have been written by any coach in any niche on any given Tuesday. That's not Claude being bad at its job. That's Claude doing its job without knowing who you are.

The fix isn't a better prompt. The fix is a better setup.

The 3 Things Claude Needs to Write Like You

Before Claude writes a single word in your voice, it needs three things.

First, it needs your custom tone & style. Not a vague instruction like "write casually and be direct." A real breakdown of how you speak, the words you use, the words you'd never use, your sentence rhythm, how you handle humor, what your opinion sounds like when you're fired up. The more specific this is, the better Claude performs.

Second, it needs examples of your actual writing. Your best captions, your email rants, your voice notes transcribed, the posts that got people sliding into your DMs. Real examples beat instructions every single time. Claude reverse-engineers your patterns from what you show it.

Third, it needs rules about what to avoid. This is the one most people skip and it's the reason their AI content still reads like AI content. Claude needs to know your specific no-go list, the phrases, structures, and formatting habits that are dead giveaways. No em dashes. No "it's not X, it's Y" flips. No three-word sentences stacked for dramatic effect. Whatever your version of AI slop looks like, name it explicitly.

How to Set This Up in Claude Without It Taking All Day

The fastest way to do this is through Claude Projects. You create a project, upload your tone guide and writing examples as files, and Claude keeps that context active for every conversation inside that project. You don't have to re-explain yourself every time you open a new chat.

If you've been starting fresh every session and pasting your instructions into the chat each time, that's why the results are inconsistent. Claude's memory doesn't carry over between chats unless you give it a permanent home to live in.

Here's the basic setup: create a project called something like "Content Voice," write out your tone guide, paste in 5-10 of your best pieces of content as examples, add your no-go rules, and test it with a caption you'd normally write yourself. Compare the output to your actual voice. Then keep refining the guide based on what's still off.

I walk through this exact process in From GPT to Claude in 5 Simple Steps , the full setup to move your entire workflow from ChatGPT to Claude and have it writing in your voice within an hour. It's $77 and you do it once.

How to Know If Claude Is Actually Writing Like You

The test is simple. Read the output out loud. If you'd never say it in a DM to a client, it's not in your voice. If a sentence could have been written by anyone in your industry, it's not specific enough. If it starts with "In today's fast-paced world" or ends with "remember, you've got this", start over.

A more ruthless test: screenshot the output and send it to someone who knows your voice well. If they don't immediately say "this sounds like you," Claude's setup isn't tight enough yet.

This is also where the AI Slop Judge earns its $22. It's a custom GPT trained specifically to call out every line that sounds generic, every structure that screams outsourced voice, every phrase that could have been written by anyone. Run your output through it before you post and it'll show you exactly what to fix.

If you want to go deeper on why AI slop is quietly costing you clients and credibility every time you post, I covered this in Episode 2 of the podcast -> Here's the invisible tax you're paying every time you post ChatGPT content. Worth a listen.

The Real Goal: Content That Sounds So Much Like You, Your Audience Doesn't Question It

When Claude is set up right, your audience shouldn't be able to tell the difference. Not because you're hiding that you use AI, but because the output is so accurate to your voice that the distinction stops mattering. You're still the one deciding what to say, what angle to take, what opinion to hold. Claude is just the one typing it faster than you can.

The coaches winning with AI aren't the ones using it to produce more volume. They're the ones who trained it well enough that it produces better content faster, in their voice, without the three-hour rewriting session that used to follow every first draft.

That's what the setup is for. And once it's done, it compounds. Every project, every conversation, every piece of content gets easier because Claude actually knows who you are.

If you're starting from GPT and want the full transition done in one sitting, From GPT to Claude in 5 Simple Steps is $77 and walks you through it start to finish.