/humanize

Overview

/humanize is a standalone command that detects and removes AI writing patterns from any text file. It scans for the telltale signs of AI-generated content — banned vocabulary, structural tells, formatting artifacts, sycophantic tone, synonym cycling — and rewrites affected passages to sound like a human wrote them. Use it for a targeted fix outside the full editorial pipeline of /dr-edit.

Usage

/humanize path/to/article.md
/humanize blog/ai-generated-draft.md
/humanize docs/api-reference.md

What It Detects

• Banned Vocabulary — words and phrases that signal AI generation: "utilize", "leverage", "delve", "it's important to note", "in conclusion", "game-changer", "cutting-edge", and dozens more.
• Structural Tells — list-heavy writing where paragraphs would be better, mechanical section transitions, formulaic introductions and conclusions.
• Formatting Artifacts — excessive bold, unnecessary subheadings for short content, over-structured layout that no human would produce.
• Communication Tells — chatbot artifacts ("As an AI...", "I'd be happy to help"), sycophantic tone ("Great question!"), knowledge cutoff disclaimers.
• Linguistic Patterns — copula avoidance ("This represents..." instead of "This is..."), synonym cycling (using a different word for the same thing each time), significance inflation ("revolutionary", "groundbreaking").
• Russian-Specific Patterns — textbook tone, restated ideas dressed up as new points, generic phrases that add no information, absurd metaphors.

How It Fixes

Three passes over the text:

1. Vocabulary and Formatting — replaces banned words, removes excessive formatting
2. Structure and Rhythm — varies sentence length, converts unnecessary lists to prose, fixes transitions
3. Anti-AI Audit — final pass to catch anything that still reads as machine-generated

Arguments

The file path to humanize. Required.

Output

Report listing all AI patterns found by category with counts. Corrected file (after approval).

Example Session

> /humanize docs/getting-started.md

Scanning for AI patterns...

Found 11 issues:
  Banned vocabulary: 4
    - "utilize" (3 occurrences) → "use"
    - "leverage" (1) → "use" / "build on"
  Structural tells: 3
    - 2 list-heavy sections converted to prose
    - 1 formulaic conclusion rewritten
  Formatting: 2
    - Excessive bold removed in 2 paragraphs
  Linguistic: 2
    - 1 synonym cycling instance normalized
    - 1 significance inflation toned down

Apply changes? (y/n)

When to Use

• Quick cleanup of AI-generated text before sharing
• Making documentation sound natural
• Preparing social media posts
• When you need AI pattern removal without fact-checking or editorial polish

Related Commands

• /dr-edit — full editorial review (includes humanize plus fact-checking and polish)
• /factcheck — companion command for fact verification
• /dr-write — create content that avoids AI patterns from the start