How to Make AI Sound Like You (Not Like a Robot)
Jack Amin
Digital Marketing & AI Automation Specialist

Quick Answer
AI defaults to a generic, over-enthusiastic writing style because it averages everything it was trained on. The fix is voice training: give AI examples of your real writing, ask it to analyse your style, and save that analysis as a reusable instruction. Once configured, the same AI that produced bland, interchangeable content will produce drafts that sound like you actually wrote them. Setup takes 15 minutes and transforms every interaction after that.
You asked ChatGPT to write a blog post. It came back with something that opens with "In today's ever-evolving digital landscape" and ends with "By implementing these strategies, you can unlock the full potential of your business."
It's grammatically perfect. It says nothing. And it sounds like every other piece of AI content on the internet.
This is the number one frustration people have with AI-generated writing. Not that it's wrong — it's that it's generic. It has no personality, no point of view, no evidence that a real person with real experience was involved.
The good news: this is completely fixable. The problem isn't the AI — it's that you haven't taught it how you sound. Once you do, the difference is dramatic.
This guide walks you through the exact process for training AI to match your writing voice — step by step, with prompts you can copy and use today. It works with ChatGPT, Claude, and Gemini.
Why does AI writing sound the same for everyone?
AI models are trained on the internet — billions of pages of blogs, articles, emails, and documents. When you give a vague instruction like "write a blog post about marketing," the AI produces the statistical average of every marketing blog post it's ever seen.
That average is:
- Slightly too enthusiastic
- Full of filler phrases nobody would actually say
- Structured identically to thousands of other articles
- Missing any specific examples, opinions, or personality
- Hedging every claim to avoid being wrong
The output isn't bad. It's just nobody. It doesn't sound like you, your business, or anyone in particular. And that's the problem — because your customers can tell.
The fix is surprisingly simple: instead of telling AI what to write, show it how you write. Then it has a real style to match instead of defaulting to the internet average.
The voice training process (15 minutes, once)
This is a one-time setup. Once you've done it, you can reuse the output across every future AI interaction.
Step 1: Collect 3–5 examples of your best writing (5 minutes)
Find pieces you've written that genuinely represent how you communicate. These could be:
- Emails to clients (especially ones where you explained something clearly)
- LinkedIn posts or social media content
- Blog posts or articles
- Proposals or pitch documents
- Long messages to colleagues or partners
- Even voice transcripts from presentations or meetings
The key is authenticity. Pick writing that sounds like you on a good day — clear, confident, and natural. Avoid anything you already had AI help write, since that defeats the purpose.
How much text? Aim for at least 500–1,000 words total across your 3–5 examples. More is better, but even 3 short emails give AI enough to work with.
Step 2: Ask AI to analyse your style (3 minutes)
Paste your examples into ChatGPT or Claude with this prompt:
Prompt: "Here are 3 examples of my writing. Analyse my writing style in detail. Specifically describe: (1) average sentence length and rhythm, (2) tone and formality level, (3) vocabulary — do I use simple or complex words?, (4) how I use examples and evidence, (5) how I structure arguments or explanations, (6) phrases or patterns I tend to use, (7) things I seem to avoid (jargon, hedging, filler, etc.), (8) overall voice — if you had to describe how I sound to someone, what would you say? Give me a concise style summary I can reuse as an instruction."
Then paste your 3–5 writing examples below the prompt.
Step 3: Review and refine the analysis (2 minutes)
The AI will produce a style summary — something like:
"Your writing is direct and conversational, with short-to-medium sentences (10–18 words average). You avoid jargon and filler phrases. You use specific numbers and real examples rather than generalisations. Your tone is confident but not aggressive — you state opinions as facts without hedging. You favour plain Australian English. You open with the most important point, not background context. You use tables and structured comparisons frequently. You rarely use exclamation marks or emojis."
Read it. Does it sound like you? If something's off — maybe it says you're formal when you're actually casual, or it missed a key characteristic — tell the AI to adjust:
"That's mostly right, but I'm more informal than you described. I sometimes use humour — dry, understated, not over-the-top. Also, I tend to address the reader directly as 'you' rather than talking about 'businesses' in the third person. Update the summary."
Step 4: Save your style profile (2 minutes)
Once the summary is accurate, save it somewhere you can easily access — a Google Doc, a note app, or directly inside your AI tool's memory system.
Now, every time you ask AI to write something, include this instruction at the start of your prompt:
"Write in this style: [paste your style summary]"
Or, better yet, configure it to be automatic — which brings us to the next step.
Step 5: Set it up so you never have to repeat this (3 minutes)
Each major AI platform has a way to save your style so it applies automatically to every conversation:
Claude → Projects Create a Project in Claude. Upload your style summary as a project instruction. Every conversation inside that Project will follow your style without you having to paste it each time. You can also upload reference documents — your service descriptions, pricing, brand guidelines — so Claude always has your business context.
ChatGPT → Custom Instructions or Memory Go to Settings → Personalisation → Custom Instructions. Paste your style summary into the "How would you like ChatGPT to respond?" field. This applies to every new conversation. Alternatively, tell ChatGPT in a conversation: "Remember that I write in this style: [summary]" and it will save it to Memory.
Gemini → Gems Create a Gem in Gemini with your style summary and business context. Use that Gem whenever you need writing that sounds like you.
Once configured, you're done. Every future interaction starts with the AI already knowing your voice, your business, and your preferences. The difference in output quality is immediate.
Before and after: what voice training actually changes
Here's a real comparison — same brief, same AI tool, same model. Only difference: one has voice training, one doesn't.
Brief: "Write a 100-word paragraph introducing a blog post about why small businesses should invest in SEO."
Without voice training (generic)
"In today's competitive digital landscape, search engine optimisation (SEO) has become an essential investment for small businesses looking to grow their online presence. With more consumers turning to search engines to find products and services, having a strong SEO strategy can significantly impact your visibility, credibility, and bottom line. In this comprehensive guide, we'll explore the key reasons why SEO should be a priority for your small business in 2026, and provide actionable strategies you can implement to start seeing results."
With voice training (in my style)
"Most small businesses treat their website like a digital business card — it exists, it looks decent, and nobody visits it. SEO changes that equation. When your site ranks for the questions your customers are already Googling, you get leads without paying for every click. It's not fast — expect 3–6 months before it moves the needle — but it compounds in a way that paid advertising never does. Here's what actually matters for SEO in 2026, and what you can safely ignore."
Same topic. Same AI. Completely different output. The second version has a point of view, a specific audience, and a voice that sounds like a person — not a content mill.
How to use your voice profile for different content types
Voice training doesn't mean every piece of content sounds identical. You adjust the instruction based on what you're writing:
| Content type | Additional instruction to add | Example tweak |
|---|---|---|
| Blog post | "Use my standard voice — conversational, direct, with specific examples" | No change needed |
| Client email | "Slightly more formal than my blog voice — professional but warm" | "Tone: professional but approachable. No slang." |
| LinkedIn post | "Shorter sentences, more personal, one key insight per post" | "Write as a LinkedIn post — under 150 words, personal tone, end with a question" |
| Proposal | "Confident and structured, evidence-based, focus on outcomes" | "Tone: business-confident. Include specific deliverables and timelines." |
| Social media | "Casual, punchy, skimmable — like a text to a friend" | "Keep it under 80 words. No hashtags in the body." |
| Technical doc | "Clear and precise, plain language, step-by-step" | "Write for a non-technical reader. Define any jargon." |
The style summary is your baseline. The content-type instruction is the adjustment. Together, they give you consistent voice with appropriate flexibility.
Advanced techniques for even better results
Once you've nailed the basics, these techniques push the quality further:
Feed AI your best-performing content
If you have blog posts, emails, or social posts that performed well (high engagement, positive feedback, generated leads), feed those specifically to the AI:
"Here are 3 pieces of content that resonated most with my audience: [paste them]. Analyse what makes these effective — structure, hooks, tone shifts, specificity — and apply those patterns to the content you write for me."
This trains the AI not just on your voice, but on your most effective voice.
Give AI your "never do" list
Sometimes telling AI what to avoid is more effective than telling it what to do:
"Never use these words or phrases in my content: leverage, synergy, cutting-edge, game-changer, dive in, it's important to note, in today's landscape, unlock your potential, revolutionise. Never open with a definition. Never end with 'in conclusion.' Never use more than one exclamation mark per piece."
Add this to your Custom Instructions or Project settings. It prevents the most common AI writing clichés permanently.
Use the "draft then humanise" workflow
For important content, use a two-pass approach:
Pass 1 — AI drafts: Give AI the topic, structure, and key points. Let it produce a complete first draft in your voice.
Pass 2 — You humanise: Read through and add the things only you can add:
- A specific story or example from your real experience
- An opinion or recommendation based on something you've seen firsthand
- A detail that proves you actually do this work (a specific tool, a specific client result, a specific number)
- A sentence that contradicts conventional wisdom based on what you've learned
Those human additions — even if they're just 3–4 sentences scattered throughout the piece — are what transform AI-assisted content into content that builds trust and authority.
Test your voice with the "who wrote this?" check
After producing a piece of AI-assisted content, show it to a colleague or friend who knows your communication style. Ask: "Does this sound like me?" If they hesitate or say "sort of," the voice training needs refinement. If they say "yes, definitely" — you've nailed it.
What to do if the AI keeps reverting to generic
Even with voice training, AI sometimes drifts back toward its default patterns — especially in longer pieces or when the topic is one it has strong pre-existing patterns for (like marketing, technology, or business).
If a section sounds generic: Copy the specific section and prompt: "Rewrite this section in my voice — more direct, more specific, less like a generic blog. Here's a reminder of my style: [paste summary]."
If the whole piece drifts: Start a new conversation (or new thread within a Project) and re-include your style instructions at the top. Long conversations can dilute initial instructions.
If it keeps using words you've banned: Be more explicit: "You used the word 'leverage' again. I've asked you not to use this word. Rewrite the sentence in plain language." AI responds well to direct correction.
If the output is consistently mediocre: Your style examples might not be strong enough. Replace them with your best writing — the emails that got great responses, the posts that performed well, the proposals that won work. Better input produces better output.
Key takeaways
- AI defaults to generic writing because it averages everything it was trained on — the fix is showing it how you specifically sound
- The voice training process takes 15 minutes once and transforms every AI interaction after that
- Collect 3–5 examples of your real writing, ask AI to analyse your style, review the summary, and save it as a reusable instruction
- Configure your style permanently in Claude Projects, ChatGPT Custom Instructions, or Gemini Gems so you never repeat the setup
- Adjust per content type — your blog voice, email voice, and proposal voice should be variations of the same baseline, not completely different
- Add a "never do" list to prevent the most common AI writing clichés permanently
- The "draft then humanise" workflow produces the best results: AI creates the structure and first draft, you add the specific details and experiences that only you can provide
- The test: show your AI-assisted content to someone who knows you and ask "does this sound like me?"
Frequently Asked Questions
Let's discuss your project
Want help with this? Get in touch.


