How to Train AI to Sound Like You in 5 Steps

In my early days of using AI as a thought partner and assistant, I remember opening a draft I had asked it to help me polish. After reading the first two lines, I knew immediately something was wrong. The grammar was clean, and the structure was tight, but none of it sounded like me.

That happened more times than I want to admit. For months, I struggled with AI-generated content because every draft came back polished and grammatically correct. But also completely hollow and sounded nothing like me.

The best way to make AI sound like you is to train it on how you actually communicate. Give it samples of your speech and writing, teach it the patterns you use and avoid, then refine the output until it matches your sound.

Why Does AI Writing Sound So Generic?

The sentences were too punchy, the transitions felt mechanical, and phrases like “this changes everything” and words like “delve” and “unleash” appeared everywhere as your writing is peppered with em-dashes and hyphens. I could spot the AI patterns immediately, which meant my audience could too.

MIT Sloan research adds a layer to this that is often missed. When people didn't know whether content came from AI or a human, they often preferred the AI version. But the moment they learned the source, human favoritism kicked in, and they rated human-made content more highly, even though the underlying quality had not changed at all. Detection is not really the risk. Once readers know who is behind the words, they reward trust in the person, not just polish in the writing.

Readers can feel the difference between a voice that is genuinely present and one that only sounds present, the same instinct I explored in Every Voice Is Valued.

Why Is Your Natural Voice Important to AI?

My breakthrough and Aha moments came when I stopped typing prompts and started speaking my ideas. I began using the AI tool Wispr Flow, a voice dictation tool that captures my thoughts as I speak, with no keyboard or editing.  I simply speak, and Wispr Flow does the rest by drafting emails, texts, documents, or anything else I'd normally type. For me, my thoughts flow much better when I am speaking instead of typing.

What I noticed was my spoken words sounded like just me, while my typed prompts didn’t as much. The difference was rhythm. When I speak, I typically use short sentences followed by longer explanations. I pause for emphasis. When I typed, it felt kind of generic and formal.  After a month of using Wispr Flow, I took that saved voice content, copied and pasted it into a Word document, and then loaded it into my AI tool. From there, it references that content when assisting me with writing so that it sounds like me.

How Do You Make AI Sound Like You?

After months of refinement, I developed a system that makes AI sound like me. It will never be perfect, and it will always be a work in progress, which is why keeping this content updated is important. Here’s what I do.

Step 1: Speak First and Edit Second

Your authentic voice lives in how you talk, not how you type.

Before writing any content, I open Wispr Flow and speak my ideas as if I’m teaching a MasterClass or coaching a leader one-on-one. There are no scripts or bullet points. I’m simply talking in my voice, tone, and style.

The transcript becomes my raw and unedited content that contains my natural rhythm, my actual phrases, and my real thought patterns. I pull from Keynote, workshop, and MasterClass transcripts, and the unscripted moments where I was teaching, not performing for a page.

Not everyone wants to talk to their phone, and that’s fine. The written equivalent works just as well. Open a blank document and free-write for 10 minutes on the same idea with no outlining or polishing as you go. You can also pull from writing you have already done, whether an email, a text to a colleague, a journal entry, or any kind of unedited writing. That unedited material carries your true voice the same way a spoken transcript does.

Step 2: Train AI on Your Real Writing

AI doesn’t know your voice until you teach it.

I collected my transcripts, articles, and emails. Then I fed them to Claude with this instruction: “Analyze my writing style. What patterns do you see in my sentence length, transitions, tone, and vocabulary? Create a comprehensive style guide based on your analysis.”

The result was a document that captured how I actually write: Confident, encouraging, short paragraphs, flowing transitions, specific numbers, and signature phrases, etc.

Step 3: Create a “Never Say” List

AI has default patterns that sound robotic, so I identified them and banned them. This is a must. My “Never Say” short list includes:

→ “It is important to note that”

→ “Certainly”

→ “I’d be happy to help”

→ “Great question”

→ “Not X. Not Y. Z.” (staccato fragments)

→ “Delve” “Unleash” "Unlock"

→ “Navigate”

→ “Leverage” as a verb

→ “This changes everything”

→ “Why this matters” or equivalent

→ Absolute “Nobody…” statements

→ Absolute “Everybody…” statements

→ Never use em-dashes and hyphens…ever!

Every time AI uses one of these phrases, I flag it and correct it. Over time, AI learns to avoid them. This, by no means, is a comprehensive list, but these are the AI-ish terms that stand out to me. There are many, many more!

Step 4: Create an “Always Sound Like” Guide

Beyond what to avoid, AI needs to know what to embrace.

My guide includes:

→ Short paragraphs (1 to 2 sentences, sometimes 3 to 4)

→ Flowing prose with natural connectors, not staccato fragments

→ Contractions when natural (don’t, can’t, it’s)

→ Signature phrases (“Here is what I have learned in my leadership journey,” “Lean in on this,” “Let that sink in”)

→ Numbers as digits (20%, not twenty percent)

→ “Just as” instead of “Similar to”

→ Connect citations back to the topic

Step 5: Iterate Until It No Longer Sounds AI-ish

The ultimate goal is for your AI-assisted writing to sound like you. What I must emphasize is that AI should assist you, not replace you. After every draft, I read it aloud, and if any sentence sounds AI-ish, I revise it. If I wouldn’t say it in my normal communications, it doesn’t belong in my content.

Common Mistakes When Training AI on Your Voice

One common mistake is training AI only on polished writing. The problem is that polished writing may not always sound like the true you. Those samples may reflect what an editor helped shape, not how you naturally communicate. Another is giving AI too little material and expecting one prompt to do the work. Your voice is a pattern, and patterns need enough examples to become visible.

Leaders also make the mistake of focusing only on vocabulary while ignoring rhythm, sentence length, transitions, emphasis, and your personal sound. Your voice continues to evolve, so you should guide AI to evolve with it. The goal is not to train AI once. The goal is to keep refining it until the output consistently sounds like you. For example, today I updated my Wispr Flow, which included the last months of talking.

Your Voice Is Your Competitive Advantage

Leaders who communicate authentically build deeper trust than those who default to corporate speak.  I broke this down in The $1.2 Trillion Communication Problem, and audiences notice the difference immediately, even when they cannot articulate why. Your voice is your competitive advantage. Only you can be you!

When you outsource your voice to generic AI output, you lose that advantage.

When you train AI to amplify your voice, you gain speed without sacrificing authenticity.

Here is the reality: AI is a mirror. If you feed it generic prompts, you get generic output. If you feed it your voice, you will get your voice back in an accelerated fashion.

The Significance of This Approach

Everything in this methodology works just as well if you never publish a single article, and if you're not careful, here is the part you might miss. You don't need to schedule a special recording session to capture your voice; you do it every time you lead a staff meeting, run a discussion, brief your team, or sit in a one-on-one.

As I already mentioned, I capture my voice intentionally with Wispr Flow. You are likely already capturing yours. If Zoom, Teams, or Otter AI is transcribing your meetings, that transcript is your raw material, and it is arguably richer than a solo recording, because it captures how you actually lead. It captures how you open a meeting, how you respond when someone pushes back, and how you close and assign next steps. Pull it from what already exists. Take the transcript from the next staff meeting, discussion, or one-on-one you lead, and feed it to Claude using the same style guide prompt I shared earlier. You don’t need to record anything new unless you have nothing.

Create a style guide document that includes.

• Email: paste five to ten of your own real emails into the AI tool of your choice and ask it to draft your next difficult email in that same voice, not a generic professional tone.

• Meeting follow-ups: use that same meeting transcript to draft the recap in your natural language instead of corporate speak, since the voice is already sitting right there in what you said.

• Performance feedback: use the same style transfer approach for written reviews, so your natural tone survives even when you are writing under deadline pressure, the exact moments when people are most likely to go silent instead of speaking honestly, which I unpack in The LIFT Effect.

• Quick messages: even a two-sentence Slack reply or Teams chat benefits from this. Ask AI to match your normal tone instead of defaulting to something that reads as if it came from anyone.

Make it permanent and save your style guide as a custom instruction or project so it applies automatically every time, not just the one time you remembered to paste it in.

Elevate With AI | Use AI as an Advisor and Assistant

Use this prompt with Claude or ChatGPT as your Advisor:

“I’m going to paste three samples of my writing. Analyze my style and create a comprehensive guide that includes my average sentence length, my transition patterns, my signature phrases, words I use frequently, words I never use, and my paragraph structure.”

Then use this prompt as your Assistant:

“Using the style guide we just created, draft my next [email, update, post, review] so it sounds like me, not like a generic AI wrote it. Flag anything you were unsure about instead of guessing.”

Coaching for Liftoff

This week, take two actions.

Action 1: Pull the transcript from the next staff meeting, discussion, or one-on-one you lead. If you do not already have one, record yourself explaining one idea for 3 to 5 minutes using Wispr Flow, your phone’s voice memo, Otter AI, or the transcribing app of your choice. Read it back, and that is your authentic voice. Notice the rhythm, the phrases, the way you emphasize certain points as well as your overall sound and patterns.

Action 2: Feed that transcript to Claude or the AI tool of your choice and ask it to create a style guide based on your patterns. Save that guide. Use it for every piece of content you create going forward.

If you do this consistently for 30 days, you will have an AI collaborator that sounds like you, not a machine that gives you AI-ish sounding outputs.

The Reality of Authentic AI

Just as one generic prompt produces hollow output, one intentional voice guide elevates every piece of content that follows. That is the reality of authentic AI.

Your audience can tell the difference between content that sounds like you and content that sounds generic and AI-produced. Give them the real you, and it will

elevate your impact, which starts with your voice.

More Ways to Elevate

If this resonated, explore these related resources. The links were provided in this article:

• The $1.2 Trillion Communication Problem: why clarity is the foundation of effective leadership communication.

• The LIFT Effect: what happens when trust breaks down, and how to rebuild it.

• Every Voice Is Valued: why people only speak up, and speak honestly, when they trust the room.

Questions Leaders Ask

How do I train AI to sound like me? Feed AI samples of your real writing, including transcripts, articles, and emails. Ask it to analyze your patterns and create a style guide. Then use that guide for every piece of content you create, and revisit it every few months as your voice keeps evolving.

What tools can help capture my authentic voice? Wispr Flow and Otter AI are excellent for voice dictation. Speaking your ideas first, rather than typing them, captures your natural rhythm and phrases.

What phrases should I tell AI to avoid? Common AI patterns include “It is important to note that,” “Certainly,” “I’d be happy to help,” and staccato fragments like “Not X. Not Y. Z.” Create a “Never Say” list and train AI to avoid these, then update it as you catch new patterns creeping back in.

How long does it take to train AI on my voice? With intentional effort, you can create a functional style guide in one session. Refinement happens over 30 days as you catch patterns and correct them, and it works well beyond content creation, in email, feedback, and everyday messages too.

LeaderLift equips leaders to learn how to make AI sound like themselves. Our hands-on approach helps leaders to apply the approach mentioned in this article so they can use AI in their authentic voice and turn that into measurable performance.

BOOK A CALL