Writing is the first job most owners give AI. It’s also where the first disappointment shows up.
Owners ask about it online in a lot of different ways:
“Why does my AI writing keep sounding generic?”
“Does using AI to write your brand story make you sound more like your competitors?”
The second one is the sharper version. A small business wins by being specific and sounding like a real person. If AI is quietly pulling your writing toward the industry average, that’s a business problem, not fussiness.
Here’s the good news. This is one of the most fixable problems in AI. It costs nothing and takes about an hour.
Why it happens
An AI writing tool learned from an enormous pile of text. When you ask it for a follow-up email, it gives you the most likely follow-up email. Not the best one. The average one.
That average is pulled from millions of businesses that all describe themselves as friendly and professional. So that’s what comes back.
This isn’t a glitch you can complain your way out of. It’s how the tool works. Ask two different AI tools the same question and you’ll get two answers in the same voice.
Which means the whole job is giving it something specific to copy. Your writing, not everyone’s.
1. Show it your writing instead of describing it
“Write in a friendly, professional tone” is the prompt most people use. It produces exactly the writing they’re complaining about, because every business on earth claims to be friendly and professional. Adjectives don’t narrow anything down.
Paste in three or four things you actually wrote instead:
- An email to a customer that went well
- A quote or estimate with your own notes in it
- A social post people replied to
- A text you sent explaining a delay
Then say: “Here are four things I wrote. Match how I write. Now draft a follow-up for this quote.”
Most owners have never tried this. It moves the output further than any amount of instruction.
What that looks like
Here’s a before and after on a job most businesses do every week: chasing a quote that went quiet. The business below is made up.
The lazy prompt: “Write a friendly, professional follow-up email to a customer who hasn’t replied to my landscaping quote.”
Subject: Following Up on Your Landscaping Quote
Hi Sarah,
I hope this email finds you well. I wanted to follow up regarding the quote we provided for your landscaping project. We understand that choosing the right contractor is an important decision, and we’re here to answer any questions you may have.
Our team is committed to delivering exceptional results, and we would love the opportunity to work with you. Please don’t hesitate to reach out if you need any additional information.
Looking forward to hearing from you!
Warm regards, The Team
Nothing is wrong with it. That’s the problem. Sarah has received this email forty times.
The same job, after pasting in four of the owner’s own emails:
Subject: That patio quote
Hi Sarah,
Just checking in on the patio quote I sent over on the 3rd. No rush.
One thing worth knowing. I’ve got a gap the week of the 24th, and after that I’m into December, which gets wet. If you want it done this year, that week is the one.
Happy to walk the yard again if you’d rather change the layout or the stone.
Dave
The second one isn’t clever. It’s shorter, it has a date in it, and it gives Sarah a reason to answer this week.
Notice where that came from. The AI didn’t know about the gap in the schedule or the wet December. The owner knew those and said so. The AI just arranged them and matched his usual sign-off. That’s the split that makes this work.
2. Write your voice down once
Samples get you most of the way. A short written spec gets you the rest, and you only write it once.
One page, six lines:
- What we call our work. “We say yard, not property. Quote, not proposal.”
- Words we never use. The ones that make you wince. Solutions. Passionate. Journey. Reach out.
- Sentence length. Short and clipped, or longer and chatty. Pick one and say it.
- How formal we are with customers. First names? “Hi” or “Dear”?
- What we never claim. No “best in town” if you can’t back it up. No guarantees you don’t actually give.
- How we sign off. The exact words.
Paste it at the top of a chat, or save it as a custom instruction so it applies to everything automatically. ChatGPT and Claude both let you do that in settings.
This is the highest-value hour in this whole article. It improves every piece of writing you produce after it.
3. Keep the last pass yourself
Not a proofread. Three specific cuts.
Delete the first paragraph. AI almost always clears its throat before it starts. “I hope this email finds you well.” “In today’s fast-paced world.” The real message usually begins in paragraph two.
Delete the last paragraph. AI likes to finish by summarizing what it just said. People don’t do that. End on the thing you want the reader to do.
Replace one generic claim with a real detail. “We pride ourselves on quality work” becomes “we used the same stone on the house two doors down.” A place, a number, a name, a date. One real detail beats a paragraph of adjectives.
The tells, so you can spot them in your own drafts
- An opener that warms up instead of starting
- “Not just X, but Y.” AI loves a balanced pair.
- Everything arriving in threes
- Over-politeness. “Don’t hesitate.” “Rest assured.” “We’d be delighted.”
- Words you’d never say out loud. Leverage. Utilize. Robust. Seamless. Streamline.
- Em dashes everywhere. That long dash nobody reaches for when texting a customer.
- A sentence that insists something is important without saying anything about it
Then read it out loud. If you wouldn’t say it to a customer standing in front of you, cut it.
Will customers be able to tell?
Probably not from the writing itself. People aren’t reliable at spotting AI text, and the tools that claim to detect it aren’t reliable either.
What customers do notice is emptiness. A message with no date, no detail, and no specific next step reads like a form letter whether a person or a machine wrote it. Your customers have had hundreds of those. What they haven’t had is one that mentions the gap in your schedule next week.
So the real question isn’t “did a machine touch this.” It’s “is there anything in here that only this business could have written.”
The short version: AI writes the draft. You add the one thing only you know. That’s what makes it sound like you.