How do LinkedIn ghostwriters capture your voice? The method, published

Voice capture is transcript work, not a questionnaire

Voice capture is the process of recording how a person actually speaks, marking the transcript for the features that repeat, and turning those marks into a document a writer consults before every draft. It is archaeology on a recording. Everything else described as voice capture is a survey with adjectives in it.

The difference matters because the two methods produce different artifacts. A questionnaire produces a page that says the executive is direct, warm and occasionally funny, which describes roughly forty million people and helps a writer with nothing. A marked transcript produces a list of the eleven constructions this person uses that nobody else in their field uses, each with a timestamp you can play back.

Most published descriptions of this method come from book ghostwriting, where the standard is a recorded author interview of two to three hours. Those pages describe the method in the abstract and stop before the artifact. They tell you the writer listens for metaphors and builds a voice profile. They never show the question set, never show what gets marked, and never show a finished profile. This page shows all three, then adapts them to the constraint a LinkedIn retainer actually runs under.

The book interview does not transfer, and the arithmetic shows why

A book interview is tape-poor and a monthly retainer is tape-rich, which is the opposite of what everyone assumes. The retainer's problem is not a shortage of recorded minutes. It is that the minutes get spent on the wrong layer.

Work it out with two assumptions you can check against your own transcript. Assume conversational speech runs near 140 words a minute, and assume the subject holds about 70 percent of the talk time in a well-run interview. A single 150-minute author interview then yields roughly 14,700 words of the subject speaking. A 60,000-word manuscript comes out of it, so the book writer has about 0.25 words of tape for every word published.

Now the retainer. A 40-minute founding call yields about 3,900 subject words. Eleven monthly top-ups of 20 minutes add about 1,960 each, so year one holds roughly 25,500 words of tape. Ten posts a month at 200 words is 24,000 words published. That is about 1.06 words of tape per published word, which is a little over four times the ratio a book interview runs on.

Capture economics, book interview against a first-year LinkedIn retainer. Assumes 140 words a minute and 70 percent subject talk time.

So the method has to change in one specific way. A book writer front-loads because there will never be more tape. A LinkedIn writer must protect the top-up from being converted into a content calendar meeting, because topics are cheap and register is expensive. That single discipline is the whole adaptation, and it is the thing nobody writes down.

The Personeur Capture Ladder

Capture runs in five rungs, and each rung produces a specific object rather than a feeling. If a writer cannot show you the object at the end of a rung, that rung did not happen.

The ladder is deliberately ordered so the cheapest material gets used first. A writer who opens with the interview is asking an executive to pay in minutes for something that was already sitting in a podcast feed.

The founding call question set, published

The questions are not about the person's career. They are designed to move the speaker out of presentation mode, because a person's real register appears when they are bored, annoyed, or telling a story that once had something at stake. Career questions produce the LinkedIn About section they already have.

One rule governs the whole call: never ask what someone's tone of voice is. Nobody can answer it, and the answers that come back are borrowed from a brand deck. Ask for behaviour and extract the tone yourself.

The founding call, twelve questions in running order, with what each one is actually for.

Twelve questions in forty minutes leaves roughly three minutes each, which is correct. The interviewer's job after asking is to stay quiet and to follow only one thread: whenever the speaker says something in a shape you have not heard from them before, ask for an example rather than moving on.

What gets marked up in the transcript, and with what

Six mark types, applied in one pass, in a fixed order. The order matters because idiolect marking is mechanical and opinion marking requires judgement, and doing the mechanical pass first calibrates the ear.

The markup legend and where each mark lands in the finished profile.

Then count, because counting is what makes this method falsifiable. A founding call of forty minutes yields roughly 3,900 subject words. One distinctive construction every 150 words is a low bar for a person with a formed voice, which sets a floor of about twenty-five idiolect marks. A first month of eight to ten posts needs at least five separate defensible positions, so the floor for CLAIM marks is five.

The monthly twenty minutes, and how not to waste it

Split the top-up five, ten, five, and protect the middle ten minutes above everything else. The default failure is that the whole twenty becomes a topic planning call, which feels productive and adds nothing to the profile.

The twenty-minute monthly top-up, allocated.

The middle ten is the part that generalises. A topic is worth one post. A position is worth a quarter of posts, because it can be argued from six angles, defended against an objection, illustrated with a story and then revisited when the market moves. Ten minutes spent forming one position outperforms twenty minutes spent listing eight topics, and the arithmetic is not close.

The last five minutes only works if corrections are recorded as rules rather than as edits. An edit fixes one sentence. A rule fixes a class of sentences and stops the same correction returning next month, which is the mechanism explained in the piece on drafts that do not sound like you.

How to know capture worked before the first draft ships

Run a blind attribution check, and run it before the retainer starts publishing rather than after month three. It takes twenty minutes and it converts an argument about feel into a number.

The Blind Attribution Check, scored. Chance performance is 50 percent.

What voice capture cannot do

Capture can only record a position that already exists. It cannot manufacture one, and no interview technique changes that. This is the honest limit of the whole method and the reason the second month of a retainer is where most disappointment actually lives.

The failure looks like this. The transcript is full of idiolect and short on claims. The writer, trying to be faithful, produces posts that sound exactly like the executive and say nothing anyone could disagree with. The executive reads them and reports that the drafts do not sound like him, because what he is hearing is his own vagueness in his own cadence, which is more uncomfortable than reading a stranger.

• Capture records how you say things. It does not decide what you think, and a writer who claims otherwise is selling you their opinions in your voice.
• Capture cannot publish what the hazard marks caught. If the best stories are all under agreement, the engagement needs a different content strategy, not a better interview.
• Capture ages. The idiolect list barely moves in a year, the story bank grows, and the opinion register goes stale fastest, which is why the monthly top-up exists at all.
• Capture does not survive an approval chain. Three reviewers can flatten a perfectly captured draft back into corporate register before it publishes, and the transcript is not to blame.

None of this makes the method optional. It makes the sequence matter. Get the positions first, capture the voice second, and price the work knowing which of the two you are actually buying, which is covered in the breakdown of what a LinkedIn ghostwriter costs.