Wikipedia Source for Humanizer Pattern Detection: Signs of AI Writing
Humanizer’s pattern detection system derives its rewrite-guiding cues directly from the Wikipedia article "Signs of AI writing," maintained by WikiProject AI Cleanup, which catalogs the linguistic indicators used to identify AI-generated text.
The blader/humanizer repository implements a skill-based system designed to detect and rewrite synthetic-sounding content by comparing text against established heuristic rules. At the foundation of this Wikipedia source for Humanizer's pattern detection architecture lies a specific crowdsourced reference page that enumerates the stylistic patterns commonly associated with large language model output.
The Wikipedia Source: Signs of AI Writing
Humanizer imports its detection patterns from the Wikipedia community-maintained page titled "Signs of AI writing." This resource, curated by WikiProject AI Cleanup, provides an authoritative taxonomy of linguistic cues that readers and editors use to identify potentially machine-generated prose.
The page organizes these indicators into five strength-ordered sections, ranging from subtle stylistic preferences to overt structural tells. Humanizer ingests these cues and maps them directly to its internal rewriting logic, enabling the tool to flag passages that exhibit AI-like characteristics such as generic claims, hedging language, or repetitive syntactic structures.
Implementation in Source Code
The repository explicitly references this Wikipedia source in two key documentation files that define the skill's behavior and public interface.
SKILL.md Line 374
In the primary skill definition file, SKILL.md, the Wikipedia page is cited at line 374 as the authoritative basis for the pattern catalog. This file contains the complete set of detection rules grouped according to the strength hierarchy defined by the WikiProject AI Cleanup guidelines. The patterns enumerated here drive the tool's rewrite engine, determining which sentences require humanization based on their similarity to the documented "Signs of AI writing."
README.md References
The project documentation in README.md acknowledges the Wikipedia source at line 71 within the primary references section, and again at lines 163-164 when discussing the theoretical foundation of the detection algorithm. These citations establish the provenance of the heuristic rules and direct users to the original research maintained by the Wikipedia community.
Pattern Detection in Action
Humanizer exposes its Wikipedia-derived detection logic through a command-line interface that processes files or stdin streams. The tool parses input text, compares it against the imported "Signs of AI writing" patterns, and generates rewritten output that minimizes AI-like stylistic markers.
Installing the Skill
Add Humanizer to your local project and make it available to all agents:
skills add . --agent '*'
This command installs the skill locally and registers it for use across every agent configuration.
Rewriting a Document
Process a single file to detect and rewrite AI-like passages:
humanizer rewrite path/to/document.md
Humanizer parses the file, flags sentences matching any Wikipedia "Signs of AI writing" pattern, and produces a humanized version with the stylistic markers removed.
Processing Text via Stdin
Analyze and rewrite content piped directly into the tool:
cat paragraph.txt | humanizer rewrite -
When receiving text via stdin, Humanizer displays three distinct stages: the initial rewrite, a brief critique of remaining AI-like elements, and the final cleaned-up version.
Explaining Pattern Matches
Request detailed information about why specific text triggers detection:
humanizer explain "This is a generic statement that sounds like an AI."
The command returns the specific pattern—such as "generic claim"—that caused the flag, directly linking the detection back to the Wikipedia source taxonomy.
Summary
- Humanizer's detection logic relies on the Wikipedia "Signs of AI writing" page maintained by WikiProject AI Cleanup.
- The patterns are organized into five strength-ordered sections that prioritize different types of AI linguistic tells.
- Source file
SKILL.mdat line 374 contains the explicit pattern definition referencing the Wikipedia source. - File
README.mdcites the source at lines 71 and 163-164 within its reference documentation. - The
humanizer explaincommand maps specific detections directly to the Wikipedia-derived pattern names.
Frequently Asked Questions
What specific Wikipedia page does Humanizer use for pattern detection?
Humanizer uses the Wikipedia article "Signs of AI writing" as its authoritative source for detection patterns. This page is maintained by the WikiProject AI Cleanup workgroup and provides a comprehensive list of linguistic indicators that suggest text was generated by artificial intelligence.
Where in the Humanizer codebase is the Wikipedia source referenced?
The Wikipedia source is explicitly cited in SKILL.md at line 374, where the pattern definitions are enumerated, and in README.md at lines 71 and 163-164, where the project lists its primary references and theoretical foundations.
How does Humanizer organize the patterns from the Wikipedia source?
Humanizer imports the linguistic cues from the Wikipedia page and groups them into five strength-ordered sections. This hierarchy allows the tool to prioritize_rewrite_ efforts based on the severity of the AI-like stylistic markers detected in the input text.
Can I see which specific Wikipedia pattern triggered a rewrite?
Yes. By running the humanizer explain command followed by the text in question, the tool returns the specific pattern name—such as "generic claim" or "hedging language"—that triggered the detection, directly mapping the result to the taxonomy defined in the "Signs of AI writing" Wikipedia article.
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:
curl -s "https://instagit.com/install.md" Maintain an open-source project? Get it listed too →