When an executive hands their writing to someone else, the fear is always the same one. A post goes out under your name, a board member or a customer reads it, and they can tell within a few lines that you did not write it. The fix has a name: voice capture. It means the writer documents how you communicate before drafting a word, then produces everything from that record instead of from imagination.
I do this work daily. I run content at Megaphone, I was the first CMO at Copy.ai, where we grew from zero to $2.5M ARR in under a year, and I have built my own accounts to 93k followers on X and 26k on LinkedIn writing as myself the entire time. What follows is the process I would demand from any writer before letting them near my name.
Voice is extracted, never invented
A writer cannot conjure your voice from your website and a headshot. What they can do is pull it out of material you already produce, because your voice is sitting there in every voice note, email, and offhand Slack message you send. The craft is in collecting that material deliberately and reading it closely enough to see the patterns.
That is the dividing line between writers who capture a voice and writers who approximate one. The first kind asks for inputs before they ask for a deadline. The second kind opens a blank page and guesses, and the result reads like a press release wearing your name tag. I covered why that failure happens, and how to decide what to delegate at all, in delegating founder content without sounding fake. This piece goes one level deeper into the capture process itself.
The raw material a good writer collects
The inputs matter more than the interview questions, so here is what a capture process should gather, roughly in order of value.
Recorded talking. Voice notes, podcast appearances, all-hands recordings, sales-adjacent calls where you explain your product to a stranger. Speech is where your real phrasing lives, because nobody performs grammar out loud. When I review a founder's recorded material, I am listening for the sentences they build without thinking, the way they hedge, and the moments they get specific without being asked.
Writing you produced under real stakes. Investor updates, the email you sent when something broke, a heated reply in a customer thread. Pressure strips out the corporate varnish and leaves your default register.
Your existing public posts. Useful, with a caveat. If your old posts were already written by someone else, they document the previous writer. A capture process should weight them below the raw material above.
Direct extraction. Prompts and questions designed to pull stories out of your head. What surprised you this week. What did a customer say to you, in their exact words. What position do you hold that people in your market would argue with. The answers arrive in your phrasing, which makes them double duty: content raw material and voice evidence at once.
What goes into the voice guide
At Megaphone we compile all of that into a document we call the Voice Guide, and whatever your writer calls theirs, it should contain the same layers.
Your sentence mechanics come first. Average sentence length, how you open, whether you hedge before or after a claim, how you handle disagreement. Then vocabulary, in both directions. The words you reach for constantly, and just as important, the banned list: words and constructions you would never use, the phrases that would make your cofounder laugh if they saw them under your name. Then your worldview frames, meaning the recurring logic you explain things through. In my experience every operator has a handful of these and reaches for them constantly without noticing.
The banned list is the layer most writers skip and the one that does the most protective work, because a reader who barely registers the absence of your favorite word will catch a foreign one instantly.
One test tells you whether a Voice Guide is real: could two different writers draft from it and land on recognizably similar posts? A mood board of adjectives will never pass that test. A document specific down to sentence patterns and banned words passes it, and at that point your voice has become an asset your content system owns instead of a vibe living in one writer's head.
Idea Capture keeps it yours
Capture is a starting state. Voices drift, positions sharpen, new stories accumulate, and a profile built in January goes stale by June unless fresh material keeps flowing in. So the process needs a loop, and ours is called Idea Capture: a standing weekly rhythm of prompts and voice notes that feeds the writer new raw material before every batch.
The founder side of that loop costs about an hour a week in our system. You record a few voice notes against specific prompts, then review the batch that comes back. The batch itself runs on a fixed plan of five X posts, three LinkedIn posts, and one long piece weekly, which matters for voice more than it sounds like it should. A regular schedule gives the writer a constant stream of your corrections, and corrections are the highest-density voice data there is.
On AI, since every executive asks. I use it constantly, and a good system gets about 80% of the way there with it. The remaining 20% is a human editor checking the draft against the voice guide and killing the tells, the stock hooks and the phrasing no real person says out loud. When delegated content stops sounding like the executive, the usual cause is raw AI output slipping through that gate.
Written voice and spoken presence are the same muscle
Most advice on executive voice covers the spoken version: how you carry a room, how you talk to senior leaders, the presence frameworks coaches teach. Kara Ronin's breakdown of developing executive presence is a solid entry into that side of it.
The page version runs on the same fundamentals. You write in plain declaratives and hold a clear position without apology, and you attach evidence to every claim with the timeframe and the mechanism named. If you communicate that way in a room, a capture process done well makes you read that way online, and the two reinforce each other. Someone who meets you after months of reading you should recognise you instantly.
How you know the capture worked
The first signal shows up when you read a draft. The reaction you are looking for is recognition, the sense that this is what you said, cleaned up and structured. If you keep hitting sentences that are competent writing you would never produce, the capture is incomplete, and more line edits will not repair it. The fix is more input material.
The second signal arrives later, in your replies and DMs. When content carries your real positions and stories, responses get specific. People push back on the take you staked out or reference the story you told. Generic engagement on delegated content usually means the voice on the page belongs to nobody in particular.
Voice capture is the difference between delegating production and delegating identity. Do it properly, with real inputs, a documented Voice Guide, and a weekly Idea Capture loop, and the person on the page stays you.
If you want a read on what your current feed is signaling before you bring a writer in, take the free feed diagnostic quiz at megaphonestudio.com. It takes a few minutes and shows you where your voice is landing today.