Blog

Branding psychology

How to sound human when everyone uses the same AI tools

When every brand drafts with the same AI tools, captions start to sound alike. How to keep a voice people recognise, with a before/after and a prompt to copy.

6 min read By Solomon Divyananth ·

Scroll through ten small-business posts on LinkedIn or Instagram this week and count how many open with "In today's fast-paced world", promise to "elevate your experience", or end with three hashtags and a rocket emoji. They read like they were written by the same person, and in a sense they were. Many of us now draft with the same handful of AI tools, and those tools default to the same safe phrasing.

I use AI every day and I'm not going to tell you to stop. This article is about the one thing AI can't supply on its own: a voice that sounds like you. In a feed full of identical captions, that voice is a brand asset.

Why everything is starting to sound the same

A language model writes by predicting the most likely next words. Left to its defaults, it picks phrasing that suits the widest range of readers. That is useful for a first draft and bad for standing out, because "most likely" is, by definition, what everyone else's tool produces too.

People have noticed. Merriam-Webster chose slop as its 2025 Word of the Year, defining it as "digital content of low quality that is produced usually in quantity by means of artificial intelligence." When a dictionary picks a mocking word for AI output, that tells you something about how audiences feel.

What the evidence says about AI and trust

Two pieces of evidence are worth knowing.

People can be wary of the word "AI" itself. In a 2024 study led by Mesut Cicek at Washington State University, published in the Journal of Hospitality Marketing & Management, researchers surveyed more than 1,000 US adults. Products described as using artificial intelligence were consistently less popular than identical products described without the term. According to the university's release, the researchers tested eight product and service categories and linked the drop to lower emotional trust, with a stronger effect for higher-risk purchases.

To be precise: that study was about products that use AI, not about AI-written marketing copy. But it tells us that, at least for now, many people feel less warmly towards things they associate with AI. Copy that reads as machine-made risks triggering the same reaction.

Search engines care about quality, not the tool. Google's guidance on AI-generated content, published in February 2023, says its systems reward "original, high-quality, people-first content" however it is produced. It also says that using AI to mass-produce content mainly to manipulate rankings breaks its spam policies. In Google's view, the problem is generic, unhelpful content, whoever or whatever wrote it.

Put those together and the goal is clear: use AI if it helps, but make sure what you publish carries your experience, your opinions and your way of speaking.

What a "voice" actually is

Your brand voice is the consistent personality in how you write and speak. Mailchimp's content style guide explains it well: your voice stays the same all the time, but your tone changes with the situation, the way you'd speak differently at dinner with friends than in a meeting with your boss.

Voice shows up in small, concrete choices:

  • Words you use and avoid. Do you say "customers", "clients" or "people"? Do you swear? Do you use jargon?
  • Sentence length and rhythm. Short and punchy, or long and conversational?
  • What you have opinions about. A voice with no opinions isn't really a voice.
  • What you notice. The details you mention show what you care about.
  • How you open and close. A greeting you always use, a question you always end on.

AI tools can imitate a voice if you describe it well. They can't invent one for you, because the raw material comes from your experience.

Five things AI can't supply for you

  1. Your stories. The client who changed their mind at the last minute, the mistake you made in year one. Real, specific events are the hardest thing for a generic tool to fake.
  2. Your opinions. "I think most logo refreshes are a waste of money" is a voice. "Logo refreshes have pros and cons" is a shrug.
  3. Your numbers. Your own results, prices and timelines, stated plainly and accurately.
  4. Your local and cultural detail. The street names, the festival rush, the phrases your customers use.
  5. Your taste. What you leave out. AI tends to add; editing is where you choose.

A worked example: before and after

Here is a made-up example: a caption for a small wedding photography studio, as an AI tool might write it from a vague prompt.

Before: Capturing life's most precious moments! ✨ At our studio, we believe every love story deserves to be told beautifully. Our passionate team is dedicated to creating timeless memories that you'll cherish forever. Book your consultation today and let us elevate your special day! 💍 #WeddingPhotography #LoveStory #Memories

Nothing in it is wrong. It could also belong to any of thousands of studios.

Here is the same caption after the owner adds a story from their own week, an opinion and their own way of speaking.

After: Last Saturday the groom's grandmother refused to sit for the family portrait. Too tired, she said. So we skipped it, and twenty minutes later I caught her dancing with the flower girl. That's the photo the couple framed. I'll always take a real moment over a perfect pose. If you want a photographer who'll stand back and wait for it, I've got a few dates left for December.

The second version is specific, has an opinion ("a real moment over a perfect pose"), and sounds like one person talking. Only that studio could write it, because only they were there.

A prompt to copy

If you draft with AI, give it your voice instead of letting it guess. Here is a prompt you can adapt:

You are helping me write a caption for my business. Write in my voice, described below. Use the story and details I give you. Do not add facts, numbers or claims I haven't given you. My voice: [3 to 5 traits, for example: warm, direct, a bit dry, short sentences, no exclamation marks] Words I use: [your words] Words I never use: [for example: elevate, journey, seamless, passionate, cherish] The story or detail: [what actually happened] My opinion on it: [one sentence] What I want the reader to do: [one action] Give me two short versions. Keep each under 80 words.

Then edit the result by reading it aloud. If a sentence is something you'd never say to a customer's face, rewrite it.

Try this today

  1. Write a one-page voice sheet. Three to five traits, a list of words you use, a list you never use, and two examples of past posts that sound most like you.
  2. Start a story file. A note on your phone where you jot real moments from your work as they happen. This becomes your raw material.
  3. Make a banned-phrase list. Collect the generic phrases you keep seeing (and deleting) in AI drafts. Add them to your prompts as "never use".
  4. Rewrite your last three captions. Add one real detail, one opinion and one thing only you could know to each.
  5. Read everything aloud before posting. If it doesn't sound like you talking, it isn't finished.