AI Detection skill
The GEO playbook Moonrank's GEO Agent uses to catch AI-flavored writing — uniform rhythm, hedging, listicle-itis — and reshape it into specific, natural prose that resonates with humans and survives AI-content classifiers.
Download the methodology, see how the agent audits a draft, or run it live inside the GEO Agent.
- Category
- GEO
- Agent
- GEO Agent
- Format
- ai-detection.md
- Price
- Free to download
The skill file
What is the AI Detection skill?
The AI Detection skill is the editing pass Moonrank's GEO Agent runs on every draft before it ships: a systematic hunt for the tells that make text read as machine-written — same-length sentences, hedging boilerplate, lists that exist to fill space, transitions nobody says out loud — followed by a rewrite built on specificity and first-hand detail.
It's written as a SKILL.md, a plain-language instruction format that people and AI agents can both follow. Read it to learn what actually flags a page as AI-generated — to readers and to classifiers — adapt it as your team's editing checklist, or let the GEO Agent run the audit and the rewrite for you.
Downloading the playbook is free with a Moonrank account. Applying it to every article before publication — flagging the tells, injecting the specifics, re-scoring the result — is part of how the GEO Agent ships content that performs.
See it in action
“Our last batch of blog posts feels off — a detector scored them 90% AI and time on page dropped. Can you fix them without rewriting from scratch?”
Audited 8 posts: 61 flagged passages, mostly hedging ('can potentially help') and stock transitions — 14 of 16 intros open with the same 'In today's landscape' pattern, and average sentence length varies by under 3 words across entire posts. Rewrote the worst 3: cut 22% of the word count, replaced 19 generic claims with numbers and named examples from your case studies. Detector scores dropped from 90% to under 15%; re-measure time on page in two weeks.
From input to fix list
- 1
You provide the draft
Paste the text or page URL, plus any first-hand material worth weaving in — quotes, data, real examples.
- 2
Moonrank scans for tells
The GEO Agent flags uniform rhythm, hedging boilerplate, filler lists, and generic transitions, passage by passage.
- 3
It rewrites with specificity
Hollow claims become real numbers, named examples, and varied cadence — or get cut entirely.
- 4
It re-scores the result
The rewrite is checked against classifiers and a human skim test, and any section still reading as filler goes back through.
Everything it checks
Tell detection
Flags rhythm uniformity, hedging, listicle-itis, and stock transitions with severity per passage.
Hedging strip
Replaces 'can potentially' boilerplate with claims the author would actually defend.
Specificity injection
Swaps generic statements for numbers, named examples, and first-hand detail.
Cadence rewrite
Varies sentence length and rhythm until the text passes a read-aloud test.
Classifier re-scoring
Checks the rewrite against AI-content detectors before and after.
Cut list
Identifies sections that can't be made specific and should be deleted, not polished.
When to reach for it
- Make an LLM-generated draft read like a human wrote it before it ships.
- Fix published pages that detection tools score as likely AI.
- Find out why generated content underperforms your hand-written pages.
- Build an editing checklist your whole team applies to AI-assisted drafts.
- Cut the hollow sections padding your articles to a word count.