Writing Captions in Your Own Voice, Not a Robot One
The caption is your voice, and automation should carry your voice — not replace it with a generic one:
The caption is the one part of the post that is genuinely yours. So it should sound like you — the way you talk at the counter, not the way a marketing robot writes.
Software can write your captions, and for a busy owner it should. The trick is teaching the tool your voice — so it writes captions you would actually say — rather than accepting the bland, exclamation-mark-strewn default that makes every business sound identical.
There is a real tension here. Automation is what makes posting sustainable; but automation done lazily produces “We are thrilled to share this wonderful feedback! 💫” — the generic robot voice nobody believes. The resolution is not to write every caption by hand. It is to make the automation sound like you.
Why the caption must sound like you
A caption in your real voice does something a generic one cannot: it makes the whole post feel like it comes from a person, not a brand.
“Maria came in convinced it’d be too dark. It wasn’t. Thanks Maria.” — that is a human being. “We are delighted to showcase another satisfied client’s transformation journey!” — that is a template, and everyone can tell. The customer’s testimonial is real; if the caption around it is robotic, you have wrapped something authentic in something fake, and the fake bleeds onto the real.
So the caption is not a throwaway. It is the frame around your proof, and a plastic frame cheapens what it holds. Write it the way you talk — the counter test: would you say this out loud to a customer without feeling ridiculous?
Why you should still automate it
Here is the tension, stated honestly: if writing captions is the thing that stops you posting, then having them written is what gets the post out.
Plenty of owners capture a great testimonial and then it sits unpublished because they cannot face writing a caption at eleven at night. For them, a caption written automatically is not a compromise — it is the difference between the post existing and not. And since the caption is yours, automating it breaks no rule: it is your content, and you are answerable for the result whether you typed it or a tool did.
So the answer is not “always write it by hand.” The answer is “automate it, but make the automation sound like you.”
Teaching the tool your voice
The difference between a robot caption and a you-caption is whether the tool knows how you talk.
A generic caption generator produces the exclamation-mark, “thrilled to share” default because it does not know you. But an automation that has your brand voice — your tone, your register, the fact that you are plain and warm and never use “journey” or “passion” — writes captions that sound like the counter, not the template.
That is the whole game with automated captions: the tool should learn your voice and apply it, so what comes out is what you would have said, faster. The failure is accepting the generic default; the success is a caption you read and think “yes, that’s how I’d have put it.” (This is the idea behind a brand-tone profile — the system writes in your voice because it has been told what your voice is.)
Two captions from the same testimonial
Picture a small café. A regular — imagine a woman called Anna — leaves a quick voice testimonial on her way out: the flat white is the only thing that gets her through a Monday, and the staff know her order before she reaches the till.
Run that through a tool on its default settings and you get: “We are thrilled to share Anna’s wonderful experience! ☕✨ Nothing makes us happier than serving our amazing community! #coffee #love”. Nobody in that café talks like that. It reads like a stranger wrote it about a place they have never been.
Now the same testimonial, in the owner’s actual voice: “Anna says our flat white is the only thing that gets her through a Monday. We’ll take that. (We did know her order before she asked — sorry Anna, you’re predictable.)”
Same testimonial underneath. Same tool did the writing. The only thing that changed is whether the tool knew how the owner talks — which is the whole difference between human and automated marketing landing in a single line.
The robot tells to strip out
Whether a caption is written by you tired, or by a tool on default settings, the same tells mark it as robotic — kill them:
- “We are thrilled/delighted/excited to…” — nobody says this out loud.
- “…transformation journey” / “passion” / “family” (unless you genuinely talk that way).
- Stacked exclamation marks and emoji — the visual noise of a template.
- Keyword-stuffing — “the best hair salon in Lugano for balayage colour services”.
- Talking about yourself — “another happy client proves our commitment to excellence”. That is bragging in the caption, and bragging is discounted on arrival.
Read the caption aloud. If it sounds like a brochure, it failed — rewrite it, or fix the automation’s voice, until it sounds like a person.
The caption is yours; her words are hers
The line that governs all of this, and the reason automating the caption is fine:
The caption is yours — your voice, your content, automate it, write it, change it, own it. The customer’s words are hers — never touched, not the audio, not the subtitles, on any platform, by any tool.
An automated caption is fine. An automated rewrite of the testimonial is a fraud. They are different objects, and the whole discipline is keeping them separate: your voice in the caption, her voice in the testimonial, and never the two confused. That single division is what makes automating the caption safe.
Where AI belongs, and where it stops
This connects to the bigger question of what to let AI write and what to keep human: AI may write your caption, your hashtags, your promotional copy — your content. It may never write, rewrite, or improve what a customer or you-in-free-text actually said.
The caption is squarely on the “AI may write it” side — as long as it sounds like you, not like a robot. Which is the whole point of this piece: automate the caption, but automate it in your voice.
But what if I don’t know my own voice?
A fair worry: plenty of owners have never once thought about how they “sound”, so being told to teach a tool your voice feels like being asked for something you do not have.
You have it already. It is how you talk to a customer across the counter — nothing more designed than that. The test is the same one as always: say the caption out loud. If it is something you would happily say to Anna’s face, it is your voice. If you would feel daft saying it aloud, it is the robot’s, and it goes back for a rewrite. You never need to name your voice or write a style guide. You only need to recognise it when you hear it, and reject it when you don’t.
Make the automation sound like you
Do not write every caption by hand — that is how the habit dies. But do not accept the generic robot default either — that is how your posts become interchangeable.
Automate the caption, teach the tool your voice, and read the result aloud: if it sounds like you at the counter, publish it. If it sounds like a brochure, fix it.
The caption is the one part that is yours. Make it sound like you, even when a machine writes it.
The full case for what to automate and what to keep human — the case against AI writing your posts — is the next piece.