Voice profiles that sound like you not like AI
Most AI tools sound the same because the prompt is the same. Voice profiles fix the prompt. The model trains against your writing, not against the average LinkedIn post.
Why generic AI posts fail
Plug a topic into ChatGPT. You'll get a perfectly grammatical post that opens with “In today's fast-paced world,” lists three bullet points, and lands in a narrow, recognisable register.
Your network notices. Engagement drops. You stop posting.
It's not the model's fault. The output is exactly what a generic prompt produces. Voice profiles are a way to give the model the specific raw material it needs to write like a real person, not a synthesised one. Same principle as the LinkedIn content strategy guide.
What goes into a voice profile
Four inputs, takes about 5 minutes:
- 3 to 5 example posts. Your own writing, ideally. Posts you admire if you're starting from scratch.
- Tone in a sentence. Casual. Authoritative. Dry. Witty. Warm. Technical. Pick the words that match how you actually want to sound.
- Topics you cover. A list of subject areas, the more specific the better.
- Hard rules. Words you never use. Hashtag preferences. Whether you use emojis. Whether you write in British English.
The model reads all of it before each generation. Contractions appear where you use them. Sentence lengths match yours. Opinions land at the strength you actually pitch them. The post reads like something you'd publish, not something you'd delete.
Multiple voices for multiple contexts
Basic accounts include three voice profiles. That matters if you manage content for more than one LinkedIn account, or want different tones for different topic areas. A founder often writes differently when they're wearing the CEO hat than when they're sharing a personal reflection on a Sunday evening, and consultants juggling client-facing and personal posts run into the same split. Keep them separate.
Voice improves with every post
Postbrander logs every edit you make before publishing. If you consistently shorten sentences, cut specific filler phrases, or remove emojis, the model picks up on that. The voice you get in month three is noticeably closer to yours than the voice you get in week one. The editing you do is doing double duty, it ships the current post and it improves the next one.
What voice profiles do not do
We are not pretending a voice profile turns AI into a clone of you. It will not reproduce specific anecdotes you have never mentioned, it will not invent opinions you do not hold, and it will not replace your judgement on what is worth saying. Every post still goes through you before it publishes. The profile is there to make that review faster, not to remove you from the loop, an approach James explains in more depth on his background page.
Related reading
Frequently asked questions
How many example posts do I need to train a voice profile?+
Three to five posts is the sweet spot. Fewer than three and the model does not have enough signal to generalise. More than ten and you start overfitting to a narrow style. Pick the posts that best represent how you want to sound.
Can I have more than one voice profile?+
Yes. Free accounts include one voice profile, Basic accounts include three. Useful if you manage content for more than one LinkedIn profile or want different tones for different topics.
Does Postbrander use my posts to train its AI?+
No. Your content is used only to generate posts for you. We never feed your writing into model training runs, and the content you paste into voice profile examples stays inside your account.
Will the voice profile improve over time?+
Yes. Postbrander logs the edits you make before publishing. Patterns, sentences you consistently shorten, phrases you consistently cut, words you reliably swap, feed back into future generations. The voice you get in month three is noticeably closer to yours than the voice you get in week one.