Ghost Ink
AI System · Email Placement

Out of Promotions

One passage, shaped for the platform you actually send from. GoHighLevel, Kit and Listmonk each need a different length and a different setup, and getting that wrong is worse than skipping it. In testing, treated mail landed in Promotions zero times out of forty. Untreated mail landed there twenty out of twenty.

See how it adapts
Reads your platform and adapts to it A resend goes out identical, every time Your visible email is never altered A verifier checks what actually went out Never touches your sending account
👻 Ghost Ink Inking Delivered
Gmail Inbox
Primary Social Promotions
9:04 Primary
8:52
8:11
7:46
8:47
8:20
7:58
Before. The same broadcast, filed under Promotions with the rest of the marketing. After. Treated mail landed in Promotions zero times out of forty in testing.
How it adapts

The same treatment on every platform is the mistake

GoHighLevel, Kit and Listmonk do not handle a broadcast the same way, so the setting that helps on one can hurt on another. Ghost Ink reads which one you send from and shapes the passage for it. Tell it the platform and it picks the rest.

PlatformPassage lengthPreheaderWhy it differs
GoHighLevel 1,200 - 3,500 words Blank, added for you Its broadcasts need the long form and their own preheader handling.
Kit 270 - 340 words None Kit surfaces more of the message than GoHighLevel does, so short is correct here.
Listmonk 270 - 340 words None Same short band, and template variables and links come through untouched.
Anything else 270 - 340 words None Falls short rather than guessing. Short is the safe direction to be wrong in.

Runs on stock Node, with nothing to install. Operated from your preferred coding agent.

Claude Code Codex Cursor
What changes

You already write the email. This decides where it lands

Nothing about your offer, your list or your subject line changes. The visible email your reader sees is the one you wrote, and that is checked on every single send rather than assumed.

The old way

Rewriting the subject line and hoping

  • Your platform reports the send as delivered, and it was. It just was not read.
  • Test six subject lines against an audience that never saw any of them.
  • Copy a tip from a thread that was written for a different sending platform.
  • Warm a fresh domain for three weeks and land in the same place you started.
  • Never actually open a real mailbox to see which tab the message went to.
The new system

One command per email, shaped for your platform

  • Picks the right treatment for GoHighLevel, Kit or Listmonk without you looking it up.
  • Writes fresh material for every email, so no two sends carry the same text.
  • Remembers what it wrote, so a resend of the same email goes out identical.
  • Reads an HTML file and writes an HTML file. It never logs into your sending account.
  • Tells you afterwards whether the message that shipped is the one you approved.
What's inside

Everything packed into the AI system

This isn't a prompt. It's the full kit that makes the AI treat your email correctly on the platform you actually send from, with the checks that stop it shipping something you did not approve.

Tooling

A zero-dependency command line tool that runs on stock Node. Nothing to install and nothing to configure.

The seven rules

Written out in full, including the ones that are per platform. This is where most of the value sits, and it is the part people get wrong.

Writer and reviewer

One model writes the passage in your voice, a second one judges it, and neither result reaches your email until both pass.

The verifier

Run it on the message your platform actually built and it tells you whether the copy that shipped still reads the way it did when you approved it.

What the testing showed

The real numbers, including the sample size and the parts that are not settled. You get the evidence, not a promise.