Ghostwriting at scale

Hold more clients than you can hold in your head.

You already know an LLM can write a competent post. The problem was never the first client — it’s the ninth, when every client’s voice, history and hard-won preferences live in your head and nowhere else. Narrative10X is where that knowledge lives instead.

Sign inInvited by a colleague? Your invitation link takes you straight in.

Straight answer: for one post on a blank page, ChatGPT or Claude will do fine. This is not a better writer. It is the memory, the evidence and the client relationship that a chat window cannot have — and those are what decide how many clients one writer can carry.

01

Memory

It remembers your client. A chat window doesn't.

Briefs, transcripts, brand guidelines, research, the strategy you agreed, and every correction anyone has ever made — all of it feeds the next draft, across campaigns. You can see exactly what reached the model and what didn't.

A chat thread throws all of it away when you close the tab.

02

Proof

It measures whether a draft sounds like them.

Twelve measures — sentence length, hedging, jargon, how often the writing addresses the reader — learned from the posts your client actually approved. Each new draft is scored against their own history, not a house style.

Ask an assistant to “match their voice” and it works from the adjectives you thought of that morning.

03

The loop

Your client closes the loop themselves.

One link, no password, no account to support. They approve, or send a post back with a note — and that note becomes a signal that changes what gets written next. They see their own numbers, not your working material.

ChatGPT cannot route work to your client or take their approval.

Voice check

Off voice · 25/100

A post written the way a general assistant writes it, scored against six months of this client’s approved work.

  • Jargon— more corporate than this client tolerates
  • Intensifiers— more breathless than their writing
  • Second person— addresses the reader more than they do
  • Sentence length— slightly longer than theirs
  • First person— on voice
  • Paragraph size— on voice

The same client’s own writing, held out of the comparison, scores 83.

How it runs

One pipeline, per client, that compounds

  1. 1OnboardA conversation, not a form. It asks until it actually understands the client.
  2. 2ResearchDeep research on their market, their competitors and what is being said.
  3. 3Agree a strategyChannels, cadence and angles, written down and followed.
  4. 4WriteDrafts built from everything above, scored against the client's own voice.
  5. 5ApproveYou, or the client in their portal. Every correction is remembered.

Every step writes back into what the client knows about itself, so the ninth post starts from more than the first did.

And the unglamorous half

The parts that decide whether you trust it

Nothing fails silently

If an uploaded file couldn't be read, it says so. If a file was too big for the context budget, it names it. You always know what the model actually got.

Originals are kept

The PDF your client sent is still there, readable, months later — not just a summary of it.

Roles that mean something

Admins run billing and the team. Writers do the work. Viewers read. Clients see only their own posts.

Every correction makes the next post better.

In a chat window, it makes the next post identical.

Sign in to Narrative10X