How to Ghostwrite LinkedIn Posts with Claude Code

Ghostwriting LinkedIn posts with Claude Code follows a three-phase workflow: research (pull the subject’s public presence), voice analysis (study their actual posting patterns for length, structure, and tone), and drafting (combine source material with voice data). The AI draft is intentionally imperfect. The subject’s reaction to it, editing lines and saying “I’d never say it that way,” is what produces the final post. Voice matching requires reading real posts, not guessing.

One-shot ghostwriting will never be good enough. If you can prompt an AI once and get a post that sounds exactly like someone, then that person isn’t bringing enough of themselves to their content. The humanity comes in the REACTION to the AI draft, always. The draft is the starting line, not the finish line.

That said, the research-and-draft phase can move a lot faster with Claude Code. Here’s how to do it right.

What ghostwriting LinkedIn posts actually means

Ghostwriting is writing in someone else’s voice. You’re not writing a post about them. You’re writing a post they would have written themselves, except faster and with more structure.

For LinkedIn specifically, ghostwriting usually means:

You have access to their source material (a podcast transcript, an interview, notes from a meeting)
You need to turn it into a post that sounds like them, not like AI
You’re going to show them a draft, they’re going to react, and that reaction is where the real work happens

The AI’s job is to get to a useful first draft quickly. The human’s job is to push back on it, rewrite lines, and make it true.

What voice matching means in practice

Voice matching is pattern recognition applied to writing. Before you draft anything, you need to know how this person writes.

That means looking at their actual posts and noting things like:

How long are they? 200 words or 600?
Do they use bullet points or write in paragraphs?
Do they start with a hook or just… start talking?
Do they thank people, mention specific names, cite numbers?
How formal is the language? Do they swear? Use contractions?
What topics do they return to?

You’re building a mental model of their defaults. When you draft in their voice, you’re applying those defaults to new material.

What scraping and Apify are

To analyze someone’s posting style, you need their actual posts. A word of caution: always verify your input data first. LinkedIn Scraping Returns No Data? is a real story about how one wrong character wasted hours.

Apify is a web scraping platform. It runs scripts that pull data from websites, including LinkedIn, and returns it in a structured format you can read or feed to Claude.

You don’t have to use Apify specifically. The point is: Claude Code can call tools, and some of those tools can fetch public post data. In this workflow, that’s how you get the raw material for the voice analysis step.

If you don’t want to set up scraping, you can paste a handful of their posts manually. The workflow still works. For a similar workflow on Instagram, see How to Scrape Instagram for Brand Research.

The three-phase workflow

Phase 1: Research

Start by asking Claude to pull everything publicly available about your subject.

Can you research the details for [Guest] across social platforms?

Claude will search LinkedIn, X, Instagram, and anywhere else with a public presence. It pulls their bio, recent posts, company info, and any interviews or articles where they’ve been quoted.

This gives you two things: context for the post you’re about to write, and a baseline sense of how they present themselves publicly.

Phase 2: Voice analysis

Once you have the research, go deeper on LinkedIn specifically.

Can you search through what [Guest] posts on his own LinkedIn to see if we have precedent?

Here Claude is looking for patterns. Not just what they post about, but how. Length, structure, tone, recurring phrases, how they open posts, how they close them.

This is the step most people skip. They ask the AI to “write in their voice” without giving it anything to go on. The voice analysis phase is what separates a post that sounds plausible from a post that sounds like them.

Phase 3: Drafting

Now you have source material (the transcript or interview) and voice data (their actual posts). Time to put them together.

Write it now more in [Guest's] voice from the transcript and from his prior social posts

Claude will:

Pull the key quotes and ideas from the transcript
Apply the voice patterns from the analysis phase
Draft a post that matches their length, structure, and tone
Flag patterns it noticed (for example: “he always thanks the interviewer at the top”)

The output will not be perfect. That’s fine. That’s the point.

The before/after

Here’s what the difference looks like on a real example.

Before (generic AI voice, no voice matching):

“Every tech revolution has sparked the same fear: that this one will be the one that finally displaces human workers permanently. And every time, the technology opened new doors even as it closed others. AI is no different.” [480 words, essay structure]

After (matched to the guest’s actual posting style):

“AI can be the great equalizer of employment opportunities. I sat down with Alex Dobrenko to talk about what that actually looks like on the ground…” [320 words, his actual structure]

The after version is shorter, starts with a direct claim instead of a historical windup, and picks up the interviewer’s name the way this person actually writes. None of that came from guessing. It came from reading his posts.

The draft is supposed to be wrong

When you show someone a ghostwritten draft, you want a reaction. “This doesn’t sound like me” is a useful data point. “This line is wrong but this one is close” is even better.

The goal of the first draft is to give them something to push off of. Most people don’t know how they write until they see something that isn’t quite right. The wrongness is the mechanism.

So after the AI draft, expect a round of edits. Expect them to rewrite lines. Expect them to say “I’d never say it that way.” That’s the conversation that produces the actual post.

What you need to run this

Claude Code with web access tools enabled
A transcript or source material (interview, podcast, notes)
The subject’s LinkedIn URL so Claude can pull their posts
Optionally, Apify or another scraping tool if you want structured post data

The workflow runs in a single session. Research, analysis, and drafting are sequential prompts in the same conversation, which means Claude builds context as it goes. By the time it’s drafting, it has the full picture.

Common Questions

How do you ghostwrite a LinkedIn post using AI?

Follow a three-phase workflow. First, research the subject’s public presence. Second, analyze their actual LinkedIn posts for patterns in length, structure, and tone. Third, draft using source material (transcripts, interviews) combined with the voice analysis. The draft gets refined through the subject’s feedback.

What is voice matching in AI ghostwriting?

Voice matching means analyzing a person’s actual writing to identify patterns: post length, use of bullet points vs. paragraphs, opening style, formality level, recurring topics, and whether they use names and numbers. These patterns inform the AI draft so it sounds like the subject, not like generic AI output.

Why shouldn’t AI ghostwritten posts be one-shot?

A one-shot AI post sounds like AI, not the subject. The value of ghostwriting comes from the subject reacting to a draft, pushing back on phrasing, and rewriting lines. That reaction is where the real voice emerges. If a single prompt produces a perfect match, the content wasn’t personal enough.

Can Claude Code research someone’s LinkedIn posting style?

Yes. Claude Code can search LinkedIn, pull recent posts, and analyze patterns. It identifies length, structure, recurring phrases, opening hooks, and tone. This analysis step is what separates a generic AI draft from one that actually resembles the subject’s voice.

A note from Alex: hi i’m alex - i run code for creatives. i’m a writer so i feel that it is important to say - i had claude write this piece based on my ideas and ramblings, voice notes, and teachings. the concepts were mine but the words themselves aren’t. i want to say that because its important for me to distinguish, as a writer, what is written ‘by me’ and what’s not. maybe that idea will seem insane and antiquated in a year, i’m not sure, but for now it helps me feel okay about putting stuff out there like this that a) i know is helpful and b) is not MY voice but exists within the umbrella of my business and work. If you have any thoughts or musings on this, i’d genuinely love to hear them - its an open question, all of this stuff, and my guess is as good as yours.