Question · AI & automation
How do I make AI-written LinkedIn posts sound like me?
Answered by Nathan Frater ·
The short answer
Build a voice file from transcripts of you talking. I take around ten call transcripts, pull out only my own words, and get the strongest AI model to analyse my intonation, sentence structure, wording and phrasing. Then I save that as a skill it uses every time it writes for me.
I came to this after reading back a batch of my own diary-style LinkedIn drafts. They were fine. They also felt curated, and that’s the problem. Readers can smell AI copy, and LinkedIn is already full of it.
So I had the model form a skill file called How to Speak Like Nathan. Transcripts of your calls make the best raw material because they capture how you talk, which is closer to the real you than anything you’d sit down and write. Strip out everyone else’s lines so the model only learns from you. Then get it to look at intonation as well as wording, because the way you pace a thought matters as much as the words you pick.
Once the skill exists, push the drafts further than feels comfortable. Shorter sentences. Let a small grammar slip through. A good diary post reads like a string of little tweets, paced out line by line.
Polished copy is what gives AI away. The bits you’d normally tidy up are often the bits that make it sound like a person wrote it.