We tried to learn people’s writing voice from their LinkedIn posts. Analyzed samples, tuned prompts, got output that was.. fine. In the right ballpark.
Then we interviewed people instead. And the transcripts blew the writing samples out of the water.
You don’t write like you talk (and that’s the problem)
Take someone who’s done marketing for fifteen years. Ask them to write a LinkedIn post about what’s changed in their industry. You’ll get something like:
“The landscape has shifted dramatically in recent years. What once worked, broad-reach campaigns with generic messaging, no longer resonates with today’s more discerning audiences.”
Fine. Professional. Forgettable.
Now get them on a call and ask the same question. Totally different person shows up:
“Honestly? Everything we thought we knew about reach is just.. it’s wrong now. I was talking to a client last week who was still running the same targeting from 2019, wondering why nothing worked. And I couldn’t even explain it without starting from scratch.”
One of those sounds like a person. The other sounds like content.
The person showed up when the performance pressure dropped.
It’s a self-monitoring thing
There’s a concept in linguistics called self-monitoring (I dug into the research behind this in a post on forensic linguistics and AI writing). How much you adjust your language based on what you think people expect.
Writing is high self-monitoring. You edit as you go. You pick “appropriate” words. You build toward conclusions you’ve already decided on. The rough edges that make your thinking yours get smoothed away. What’s left is the professional version of what you think. Not the actual version.
Conversation is low self-monitoring. You use words you actually reach for, not words you think you should use. You hedge when you’re unsure. You restart sentences. You say “I don’t know” and then answer anyway.
That’s where the real voice lives. In the restarts, the “honestly?”, the specific client from last week.
(This is why every good ghostwriter starts from interview transcripts, not writing samples. They’ve known this forever.)
What transcripts catch that writing doesn’t
Read any transcript carefully and you’ll find stuff that would never survive a writing process.
The weirdly specific detail. The client’s name, the exact number, the date it happened. Writing drifts toward the general. Speaking pulls toward the specific, because that’s how you make someone get it in real time.
The honest hedge. “I think this is right, but I’m not sure if..” Writing kills that. Speech keeps it. And that uncertainty is often the most interesting part of the whole idea.
The first phrase you reached for, before the “better” one. In writing you’d replace it. In speech it stays. And sometimes that first instinct was more accurate. Just less polished.
My own hesitant transcript became a sharp opening line
When I interviewed myself about leaving the company I’d built for thirteen years, this is what came out:
“I’d only been a CEO for two years. I mean, our company was still quite raw, like 15 people, 20 people, something like that..”
Not polished. Hesitant, circling. But the real thing is in there. The rawness, the human gap, the uncertainty about what I was even trying to say.
The post that came from that interview opened with:
“I quit as CEO of the company I co-founded. Twice.”
Spoken version, compressed and sharpened. The restarts are gone but the honesty stayed. The admission I quit twice, the implication I wasn’t sure I should have left. That stuff would’ve been edited out of a written draft. It survived because it was in the transcript.
The issue isn’t execution. It’s source material.
If you’ve been trying to improve your LinkedIn posts by writing better.. maybe stop? Posts that sound generic are generic at the idea level. The thought hasn’t been tested in conversation yet, hasn’t been challenged, hasn’t been sharpened by a follow-up question.
Good content starts with your actual voice. An interview transcript, a voice memo, a conversation you jotted down after. That’s where content stops being generic. Because it comes from a specific person, in a specific moment, thinking through something before they’ve had a chance to make it sound like everyone else.
The polishing is the easy part 🤷
At Outerview, every post starts from a conversation with our AI interviewer, not from a blank page. Your speaking voice becomes the foundation for everything that gets written. Start your first interview at outerview.app
