An AI voice check is a short editing pass that makes an AI-assisted draft sound like your business before a customer ever reads it. Most small businesses I work with skip it. Someone asks a chatbot for a blog post or a follow-up email and sends it after a quick skim for mistakes.
The facts in those drafts are usually fine. The voice is what gives them away. Your customers have read a lot of AI copy by now, and even when they can’t name what feels off, they notice it. The email sounds like every other email in their inbox, and the trust you built on the phone gets a little thinner.
AI models keep getting better at sounding human, and they still carry habits a careful reader picks up on. Below we walk through why those habits stick around and how to build a fifteen-minute routine that catches them before your customers do.
Why Does AI Writing Still Sound Like AI?
AI writing sounds like AI because language models lean on the phrases and patterns that showed up most often in what they learned from. They’re built to predict the most likely next word, so they drift toward safe, familiar wording. You end up with a lot of “can help you,” “plays a key role,” “unlock new possibilities,” and “whether you’re a beginner or a seasoned pro.”
Rhythm gives it away too. AI tends to write sentences of about the same length, one after another, with a tidy wrap-up at the end of every paragraph. People write in bursts. A long, winding sentence, then a short one. Sometimes a fragment.
A draft can be accurate and still feel generic. Generic copy is easy for a reader to scroll past, and that’s the real cost for a business that depends on people remembering its name.
Do Your Customers Actually Notice?
They notice more than most owners expect. Few customers could point to the exact phrase that tipped them off, but after seeing so much machine-written copy, they recognize the overall feel of it. A message that sounds templated gets read like a template: quickly, and without much attention.
The cost shows up in small ways. A follow-up email that sounds like a form letter gets fewer replies. A blog post that reads like every other post on the topic gives a reader no reason to come back or remember who wrote it.
For a small business, voice is part of what people are buying. Customers choose you partly because of how you talk to them, and every piece of copy that loses that voice spends a little of the relationship.
Does a Fixed List of AI Writing Giveaways Stop Working Over Time And Why?
A fixed list stops working because every model update shifts the habits. A pattern that was everywhere in last year’s AI writing can almost vanish in the next version, while new pet phrases show up to take its place. Teams that memorized one set of tells end up checking for problems the current model no longer has.
Different tools also have different quirks. One model loves superlatives and calls everything “the most powerful option on the market”. Another hedges every claim until the sentence barely says anything. If your team uses more than one AI tool, a single generic checklist misses a lot of what’s actually sitting in your drafts.
That’s why the routine has to learn from your own work. The useful list is the one built from what your tools are writing for you right now, and it changes as they do.
What Should You Check?
Five parts make up the routine: real samples of your own writing, a running list of phrases to cut, a rewrite prompt that ties them together, a read-aloud pass, and one person who owns it. Each piece is simple on its own. Together they turn a vague sense that something sounds off into a process anyone on your team can repeat.
The samples give the AI something concrete to aim for, and the phrase list tells it what to strip out. The prompt handles most of the rewriting. Reading aloud catches whatever slipped through, and having an owner keeps the routine from fading after the first month.
Most emails and social posts take ten to fifteen minutes to run through all five parts. A full blog post takes a bit longer, though it rarely needs more than half an hour.
Why Should You Collect Samples of Your Brand Voice?
Pull a handful of pieces you or your team wrote without AI that sound the most like your business on a good day. Emails that got great replies work well. So do proposals that closed, newsletters people replied to, social posts people commented on, and anything a customer once quoted back to you.
Aim for variety in format and consistency in voice. Four or five samples covering a couple of different situations give the AI enough to pick up your sentence length and the words you reach for. Keep them in one shared document so nobody has to hunt for them when a draft is due.
Spoken language counts too. If you’ve already captured how your best salesperson answers customer questions in an AI expertise engine, those recorded answers make excellent samples, because they show how your business sounds when it’s talking to a real person.
How Is a Tell List Built and Maintained?
Start a document and add every phrase you catch yourself deleting from AI drafts. Typical early entries include openers such as “in today’s world,” filler such as “it’s important to note,” stock phrases such as “a wide range of,” and superlatives such as “the ultimate solution.” Within a couple of weeks you’ll have a list that reflects your tools and your industry.
Add structural habits as well. If your AI keeps ending paragraphs with a one-line summary, or opens every email by restating the customer’s question, write that down. Patterns like those are harder to spot than a single phrase, and they’re often what makes a piece feel machine-made.
Review the list whenever you switch tools or your provider releases a new model. Cut entries that stopped showing up and add the new ones you’re seeing. A short list that stays current will catch more than a long one nobody has touched since spring.
How Do You Create the Rewrite Prompt?
The prompt should give the AI your samples and your tell list along with the draft, then ask for a rewrite in your voice that keeps every fact the same. Asking it to explain its edits matters as much as the rewrite itself.
Here’s a version you can adapt:
Act as an editor who knows my writing voice. Here are samples of my own writing: [PASTE SAMPLES]. Here are phrases and habits I want removed or replaced: [PASTE TELL LIST]. Here is an AI-assisted draft: [PASTE DRAFT]. Rewrite the draft so it sounds like my samples. Replace everything on my list, cut any superlative that isn’t backed by a specific fact, drop filler opening sentences, and vary sentence length the way I do. Keep every fact and claim the same. Then list each change you made so I can review it.
That change list at the end is where the value sits. It shows you exactly what moved, so you can catch a rewrite that shifted a meaning or softened a promise you meant to make. It also teaches you what your tools keep getting wrong, which feeds straight back into the tell list.
Should You Read the Draft Out Loud?
Reading aloud catches problems your eyes slide right past. When you hear a draft, you notice several sentences in a row that land at the same length, or a phrase you’d never say to a customer’s face. If you stumble over a line, a reader will too.
It also works as a final gut check on tone. Ask yourself whether you’d say this, in these words, to the person receiving it. If the answer is no, keep editing until it’s yes.
This step takes two or three minutes for an email and maybe ten for a long article. It’s about the cheapest quality control a small business can run.
Who Should Manage The Tell List?
One named person should own it, and it should run on everything that goes out under your business’s name. That covers marketing emails, blog posts, proposals, social posts, and any customer reply drafted with AI help. Internal notes and meeting summaries can skip it.
The owner keeps the samples and the tell list current and spot-checks drafts from the rest of the team. In a very small business, that’s probably you. On a team of ten or twenty, it’s often whoever already handles marketing or customer communication.
Put a short review on the calendar once a month. Look over a few recent pieces and update the tell list with anything new that slipped through. If some of your recent writing beats one of the old samples, swap it in.
How Do You Start Your First AI Voice Check This Week?
Start with the next AI-assisted piece you’re about to send. Spend twenty minutes pulling four samples and writing down the phrases you already know bother you, then run that draft through the prompt and read the result out loud. That first pass gives you a working routine you can sharpen as you go.
Bring your team in once you’ve tried it yourself. Show them the before and after, share the samples document and the tell list, and ask them to run the check on their own drafts for a week. Their edits will grow the list faster than you could alone.
AI drafts will keep improving, and their habits will keep shifting from one model to the next. An AI voice check keeps your business sounding like itself through every one of those changes, so the customers who chose you for how you talk keep hearing that same voice in every email and post.
Frequently Asked Questions
What is an AI voice check?
An AI voice check is a short editing routine that rewrites AI-assisted drafts to match your business’s real voice before they go out.
How long does it take?
Most emails and social posts take ten to fifteen minutes. A long blog post might take closer to half an hour.
Do I need special software?
No. Any AI writing tool you already use can run the rewrite prompt, and the samples and tell list can live in a shared document.
What should go on a tell list?
Any phrase or habit you keep deleting from AI drafts, such as stock openers, filler phrases, empty superlatives, and repetitive paragraph endings.
How often should the tell list be updated?
Review it once a month and any time you switch AI tools or your provider releases a new model.
Who should run the check?
One named person should own the routine, and anyone drafting customer-facing copy with AI should run it on their own drafts before sending.
Does every AI-assisted draft need an AI voice check?
Run an AI voice check on anything that goes out under your business’s name. Internal notes and meeting summaries can skip it.