Humanizer Pattern Identification: The Wikipedia Source Behind AI Detection

Humanizer derives its AI detection patterns from the Wikipedia article "Signs of AI writing," curated by WikiProject AI Cleanup.

Humanizer is an open-source skill for detecting and rewriting AI-generated prose. According to the blader/humanizer repository, the tool's pattern identification system imports characteristic phrasing and stylistic tells directly from a specific community-maintained Wikipedia page, organizing these rules into five strength-based sections to detect and sanitize machine-written text.

The Wikipedia Source: "Signs of AI Writing"

WikiProject AI Cleanup Authority

The patterns originate from the Wikipedia page "Signs of AI writing" maintained by WikiProject AI Cleanup. According to SKILL.md lines 374-376, the documentation explicitly states: "The patterns come from Wikipedia's [Signs of AI writing] maintained by WikiProject AI Cleanup..." This community-curated resource enumerates characteristic phrasing, formatting quirks, and stylistic habits typical of language-model-generated text, providing the empirical foundation for Humanizer's detection logic.

Five Strength-Based Pattern Sections

Humanizer organizes the Wikipedia-derived patterns into five strength-based sections. This taxonomy allows the skill to weight different AI tells according to their severity, enabling nuanced detection that distinguishes between subtle robotic phrasing and obvious machine-generation artifacts. The graduated approach ensures lightweight editing for minor issues while flagging heavily AI-saturated passages for comprehensive rewrites.

Implementation in the Humanizer Codebase

SKILL.md and the Source Reference

The core pattern list resides in SKILL.md, which serves as the skill's primary prompt definition. Lines 374-376 contain the explicit citation linking the detection rules to the Wikipedia source. This file defines how the skill recognizes AI-generated prose and structures the rewrite prompts to target specific tells identified by the WikiProject.

Key Repository Files

Three files define how Humanizer imports and applies the Wikipedia-derived patterns:

  • SKILL.md — Contains the core skill prompt, the complete pattern list, and the Wikipedia source reference at lines 374-376.
  • README.md — Provides installation instructions, usage examples, and citations to the "Signs of AI writing" Wikipedia page.
  • AGENTS.md — Offers guidelines for modifying the skill's behavior without breaking compatibility with agent platforms.

Using Humanizer for AI Detection

Installation

Install Humanizer globally using the Skills CLI:

npx skills add blader/humanizer --global

Pasted Mode

Run the skill interactively to rewrite text pasted directly into the terminal:

/humanizer
> It's not just about the beat riding under the vocals; it's part of the aggression and atmosphere. It's not merely a song, it's a statement.

Humanizer outputs a draft, flags remaining AI tells, and provides the final rewrite.

File Mode

Process an entire document while preserving code blocks, frontmatter, and links:

/humanizer path/to/document.md

The skill returns a concise summary of changes made to the prose.

Embedded Mode

Integrate Humanizer into automated workflows by embedding it as a skill in JSON tasks:

{
  "task": "pull-request",
  "skill": "/humanizer",
  "input": "docs/api.md"
}

In embedded mode, the skill returns only the final human-readable text, ready for inclusion in pull requests or other automated pipelines.

Summary

  • Primary Source: Humanizer's pattern identification relies entirely on the Wikipedia page "Signs of AI writing" maintained by WikiProject AI Cleanup.
  • Source Location: The Wikipedia reference is explicitly documented in SKILL.md at lines 374-376.
  • Pattern Organization: Detection rules are categorized into five strength-based sections for graduated editing.
  • Operational Modes: The skill supports pasted mode, file mode, and embedded mode for different integration scenarios.
  • Preservation Logic: File mode specifically preserves code blocks, frontmatter, and links while rewriting prose.

Frequently Asked Questions

Which Wikipedia page inspires Humanizer's pattern identification?

Humanizer imports its AI detection patterns from the Wikipedia article "Signs of AI writing." This page is maintained by WikiProject AI Cleanup and provides the characteristic phrasing and stylistic tells that the skill uses to identify machine-generated text.

How does Humanizer organize the detection patterns from Wikipedia?

Humanizer organizes the Wikipedia-derived patterns into five strength-based sections. This structure allows the tool to apply graduated editing—addressing minor stylistic issues lightly while comprehensively rewriting passages that exhibit multiple strong AI tells.

Where is the Wikipedia reference documented in the Humanizer source code?

The reference appears in SKILL.md at lines 374-376. This file serves as the core skill prompt and explicitly cites "Signs of AI writing" as the source for the pattern list, directly linking the implementation to the WikiProject AI Cleanup documentation.

Can developers modify the pattern detection in Humanizer?

Yes. Developers can modify the skill's behavior by editing SKILL.md, though AGENTS.md provides specific guidelines for making these changes without breaking compatibility with agent platforms. The modular design allows customization of pattern weights and rewrite behaviors while maintaining the core Wikipedia-derived detection logic.

Have a question about this repo?

These articles cover the highlights, but your codebase questions are specific. Give your agent direct access to the source. Share this with your agent to get started:

Share the following with your agent to get started:
curl -s "https://instagit.com/install.md"

Works with
Claude Codex Cursor VS Code OpenClaw Any MCP Client

Maintain an open-source project? Get it listed too →