The AI Tell Kill List · Self-Editing Prompt Pack

Your readers can tell.

The em dash. The “it’s not X” seesaw. The three parallel sentences building to a hollow crescendo. AI-assisted writing carries fingerprints, your sharpest readers spot them in seconds, and every one they spot costs you credibility you spent years earning.

The AI Tell Kill List is a 4-tier editing system, built and battle-tested through a daily publishing practice, packaged as three paste-into-Claude prompts. Paste one prompt and your drafts stop reading like AI.

Get the pack $29 launch price · one-time purchase

Checkout opens after final pre-launch review.

The AI can’t see its own tells. That was the test that started this.

Every single model.

In a June 2026 cross-test, the same editorial constraints went to a spread of frontier AI models, and each was asked to certify its own output as clean. Every model certified itself clean. The tells sat in plain sight.

Asking your AI “does this read like AI?” gets you reassurance. The prompts in this pack force something different: per-pattern counts and exact quotes before any verdict. In verification testing, the scanner prompt caught all 11 planted zero-tolerance tells in a seeded draft, with exact quotes and paragraph locations, and returned a BLOCKED verdict a self-check would have waved through.

What the fix looks like

Before · deliberate specimen of the tells

“It’s not the workload that breaks you — it’s the system you built to manage it. We streamline. We optimize. We eliminate. In conclusion, the tools were never the answer.”

After · same idea, human on the page

“The system you built to manage the workload is what breaks you. We dig into the calendar together and cut whatever stopped earning its place. Last spring that meant 19 hours a week back for one client, measured against her own calendar audit.”

What’s inside

Three files, each in markdown and PDF:

The Kill List (version 4.6) The full 4-tier taxonomy: 4 zero-tolerance patterns, 24 hard kills, 33 frequency rules, and platform-specific blocks for columns, LinkedIn, newsletters, and short-form. Every pattern comes with a test and a fix.
The Prompt Pack Three paste-in prompts, tested against Claude and ChatGPT: an always-on system prompt that applies the rules while the model writes, a draft scanner that returns tier-ordered findings with exact quotes, and a line rewriter that repairs flagged lines and audits its own output.
The Positive Signature Worksheet Clean and generic is still a failure. Seven verified craft principles from the Ogilvy, Zinsser, Godin, and Handley canon, plus a 12-category worksheet for extracting your own voice signature from your own writing.

Who this is for

Writers publishing AI-assisted work under their own name. Ghostwriters whose clients’ reputations ride on every line. Marketers shipping copy at volume. Founders whose LinkedIn posts are the company’s public face.

I’m Marissa Brassfield. I’ve spent 15+ years writing and editing professionally: 1,700+ published articles, national business press columns, a weekly newsletter, and ghostwriting for founders whose names carry real weight. This list is the editing floor under all of it, maintained version by version as the models and their tells evolved.

Questions, answered

Which AI tools does this work with?: Claude and ChatGPT are both tested. The prompts are plain text and run anywhere you can paste a prompt.
Is this a subscription?: One purchase. The list is versioned (you buy 4.6), and refreshed editions ship to the same download link.
Why prompts instead of software?: Because the editing standard is the product. Prompts are inspectable, tweakable, and yours; you can read every rule this pack applies and add your own.
What if I write in a niche?: The worksheet in file 03 exists for exactly that: it walks you through extracting your own positive signature so the model writes toward your voice, and the scanner prompt accepts your additions as new numbered rules.