AI captions that still sound like you
The most common complaint about AI-written captions is that they sound like everyone else’s AI-written captions. Flat, over-eager, three emoji and a rhetorical question. The reflex is to blame the model. Usually it’s the brief.
A model writes to the ceiling you set. Give it “write a post about our new feature” and you get the median of the internet. Give it your voice, your audience, and one real opinion, and it writes something you’d actually publish.
What a good brand model carries
- Voice, in examples. Not adjectives (“friendly, professional”) — actual posts you liked, so the pattern is learnable. Three good ones beat a paragraph of description.
- Audience. Who reads this, what they already know, what they’re skeptical of. Copy that assumes the reader is smart reads completely differently.
- Pillars. The three or four things you talk about, so the output stays on-topic instead of drifting to generic marketing.
- Do-not-say rules. The words, claims and clichés that are off-brand. This single list removes most of the “AI smell.”
The tell to watch for
If a draft could belong to any brand in your category, it isn’t done. The fastest fix is a specific detail only you would know — a number from your own data, a customer’s exact words, a strong take you’d defend. Specificity is what makes copy sound human, whoever or whatever wrote the first pass.
Used this way, AI isn’t a voice replacement. It’s a very fast first draft that already knows your rules — so the editing is minutes, and the result is yours.