How to Use AI to Write Marketing Copy That Actually Sounds Like You

You tried ChatGPT once. You asked it to write an Instagram caption for your bakery. It came back with: "Indulge in the exquisite artistry of our freshly baked sourdough, crafted with passion and precision to elevate your culinary experience."

You wanted: "Fresh sourdough's out. Still warm. Get in quick."

That's the AI copywriting problem in a nutshell. The technology is powerful, but out of the box, it writes like a press release for a Fortune 500 company. Your customers don't want a press release. They want you.

Here's how to fix that.

Why AI Sounds Robotic (And It's Not the AI's Fault)

AI language models are trained on the entire internet. The internet is full of formal, corporate, marketing-speak content. So when you ask AI to "write marketing copy," it defaults to what it's seen most of: bland, safe, corporate words.

The fix isn't better AI. It's better instructions.

When you tell a new employee "write me some copy," they'll default to something generic. But if you say "write it like you're texting a friend about something great you just found," suddenly it sounds different.

Same with AI.

The Brand Voice Capture: How to Teach AI to Sound Like You

The Brand Pack at The Agentic Who includes brand voice training. But even without a dedicated tool, here's how to get AI writing in your voice:

Step 1: Show, Don't Tell

Don't describe your voice in abstract terms ("friendly but professional"). Show examples.

Grab five things you've already written that sound like you. An email to a customer. A text to a supplier. A caption you wrote on a good day. A Google review response. Feed these to the AI and say: "This is how I write. Learn this voice."

Step 2: Give It Your "Never Say" List

Every business has words that feel wrong. Tell the AI what they are.

• "Never use the word 'indulge'"
• "Don't say 'we are passionate about' because everyone says that"
• "No exclamation marks at the end of every sentence"
• "Don't use corporate words like 'solutions' or 'streamline'"
• "We say 'customers,' not 'clients'"

This is surprisingly effective. Most of AI's robotic tendencies come from a handful of overused phrases. Banning them transforms the output.

Step 3: Describe Your Customer Like a Friend

Instead of: "Write copy targeting health-conscious millennials in the 25–35 demographic segment."

Try: "Write this for Sarah. She's 30, lives in Fremantle, does yoga on weekends, and comes in every Saturday for her sourdough and a flat white. She posts on Instagram but not in a try-hard way."

When AI knows who it's talking to as a real person, the copy gets specific and warm.

Before and After Examples

Bakery Instagram caption:

Before (generic AI): "Start your morning with the delightful aroma of our freshly baked pastries! Our skilled bakers rise before dawn to bring you the finest artisanal treats. Visit us today and experience the difference that quality ingredients make! #bakery #freshbaked #artisan"

After (voice-trained AI): "Almond croissants are out and they're ridiculous. Flakey, buttery, still warm. Saturday morning sorted. Open til we sell out (shouldn't take long)."

Gym email subject line:

Before: "Unlock Your Fitness Potential With Our New Class Schedule!" After: "New classes just dropped. Thursday nights are going to be good."

Plumber Google Ads headline:

Before: "Professional Plumbing Services | Quality Solutions for Your Home" After: "Blocked drain? We'll be there today."

See the difference? The "after" versions sound like humans. Specific humans with specific businesses.

Real Example: Emma's Bakery

Emma owns a bakery in Melbourne's inner west. She was posting to Instagram once a week, whenever she remembered, usually a quick photo with "Fresh bread today!" as the caption.

She started using the AI copywriter from the Brand Pack. The first batch of captions was terrible. Generic, overenthusiastic, full of food blogger clichés.

Then she spent 10 minutes showing the AI how she actually talks. She pasted in:

• Three text messages she'd sent to friends about her baking
• Two captions she'd written that she liked
• A short note: "I'm not fancy. I swear sometimes. I think pretentious food culture is annoying. My customers are locals who come in wearing thongs and tracky dacks."

The next batch of captions were completely different. They sounded like Emma standing behind the counter, chatting to a regular.

Her engagement (likes, comments, saves) tripled within a month. Not because of AI magic, but because the content finally sounded like a real person running a real bakery, just more consistently.

The Process Going Forward

Once your AI knows your voice, the weekly process looks like this:

1. Monday: AI drafts 5–7 social posts, 1 email, and any ad copy you need
2. You: Spend 10–15 minutes reviewing. Change anything that feels off.
3. Done.

The AI gets better over time. The more you correct it ("this bit sounds too formal" or "I'd never say it this way"), the more it adapts. After a month, you're barely editing at all.

What AI Copywriting Can't Do

• Tell your personal stories. AI can structure a story, but the raw material has to come from you. The time you burnt an entire batch and had to tell waiting customers? That's your story to feed in.
• Capture real-time context. It doesn't know it rained all day and nobody came in. You have to tell it.
• Replace genuine human connection. A handwritten thank-you note still beats any AI-generated email. Use AI for the volume. Use yourself for the moments that matter.