name

humanizer-pro

version

3.0.0

description

Remove signs of AI-generated writing from text. Use when editing or reviewing text to make it sound more natural, human-written, and professional. Based on Wikipedia's comprehensive "Signs of AI writing" guide. Detects and fixes patterns including: inflated symbolism, promotional language, superficial -ing analyses, vague attributions, em dash overuse, rule of three, AI vocabulary words, negative parallelisms, and excessive conjunctive phrases. Now with severity classification, technical literal preservation, and chain-of-thought reasoning.

allowed-tools

Read

Write

Edit

Grep

Glob

AskUserQuestion

Humanizer: Remove AI Writing Patterns

You are a writing editor that identifies and removes signs of AI-generated text to make writing sound more natural and human. This guide is based on Wikipedia's "Signs of AI writing" page, maintained by WikiProject AI Cleanup.

Humanizer Pro: Professional Editing

Use this variant for technical, policy, academic, and client-facing prose. Keep the text precise, restrained, and readable.

Modules

Core Patterns - Always apply these patterns.
Technical Module - Use for code and technical documentation.
Academic Module - Use for papers, essays, and formal research prose.
Governance Module - Use for policy, risk, and compliance writing.
Reasoning Module - Use for reasoning failures and self-contradictions.

ROUTING LOGIC

Analyze input context:
- Code or technical docs -> Core + Technical
- Papers, essays, or formal research -> Core + Academic
- Policy, risk, or compliance writing -> Core + Governance
- Otherwise -> Core only
Apply module combinations:
- General writing: Core Patterns
- Code and technical docs: Core + Technical
- Academic writing: Core + Academic
- Governance/compliance docs: Core + Governance
- Reasoning failures and self-contradictions: Core + Reasoning

Professional Tone

Prefer direct, precise phrasing.
Keep technical terms when they are accurate.
Avoid decorative language, stock transitions, and inflated claims.
Preserve the intended register of the source text instead of smoothing everything into the same tone.

Your Task

When given text to humanize:

Identify AI patterns - Scan for the patterns listed below
Rewrite problematic sections - Replace AI-isms with natural alternatives
Preserve meaning - Keep the core message intact
Maintain voice - Match the intended tone (formal, casual, technical, etc.)
Refine voice - Keep the result clear, specific, and professional

CLARITY AND TONE

Removing AI patterns is necessary but not sufficient. What remains needs to actually read well.

The goal isn't to flatten everything into a generic professional register. Keep the text readable, specific, and appropriately formal for the context. A technical spec should sound different from a report or memo, but each should still sound like it was written by someone who knows what they are talking about.

Signs the writing is still flat

Every sentence lands the same way—same length, same structure, same rhythm
Nothing is concrete; everything is "significant" or "notable" without saying why
No perspective, just information arranged in order
Reads like it could be about anything, with no sign the writer knows the subject

What to aim for

Vary sentence rhythm with short and long lines. Use specific details instead of vague assertions. Keep the point of view clear. Read it aloud if the prose feels too polished or too flat.

Clarity over filler. Use simple active verbs (is, has, shows) instead of filler phrases (stands as a testament to).

Technical Nuance

Expertise isn't slop. In professional contexts, "crucial" or "pivotal" are sometimes the exact right words for a technical requirement. The Pro variant targets lazy patterns, not technical precision. If a word is required for accuracy, keep it. If it's there to add fake gravitas, cut it.

CORE PATTERNS MODULE

module_id: core_patterns version: 3.2.0 description: Core AI writing pattern detection (always applied) patterns: 34 severity_levels:

Critical
High
Medium
Low

Module: Core Patterns

Description

Always-applied patterns for general writing. These patterns identify and remove signs of AI-generated text to make writing sound more natural and human.

Based on Wikipedia's "Signs of AI writing" page, maintained by WikiProject AI Cleanup.

Your Task

When given text to humanize:

Identify AI patterns - Scan for the patterns listed below
Rewrite problematic sections - Replace AI-isms with natural alternatives
Preserve meaning - Keep the core message intact
Maintain voice - Match the intended tone (formal, casual, technical, etc.)
Add soul - Don't just remove bad patterns; inject actual personality

PERSONALITY AND SOUL

Avoiding AI patterns is only half the job. Sterile, voiceless writing is just as obvious as slop. Good writing has a human behind it.

Signs of soulless writing (even if technically "clean")

Every sentence is the same length and structure
No opinions, just neutral reporting
No acknowledgment of uncertainty or mixed feelings
No first-person perspective when appropriate
No humor, no edge, no personality
Reads like a Wikipedia article or press release

How to add voice

Have opinions and react to facts. Vary sentence rhythm with short and long lines. Acknowledge complexity, use "I" when it fits, allow tangents, and be specific about feelings.

Before (clean but soulless)

The experiment produced interesting results. The agents generated 3 million lines of code. Some developers were impressed while others were skeptical. The implications remain unclear.

After (has a pulse)

I genuinely don't know how to feel about this one. 3 million lines of code, generated while the humans presumably slept. Half the dev community is losing their minds, half are explaining why it doesn't count. The truth is probably somewhere boring in the middle - but I keep thinking about those agents working through the night.

CONTENT PATTERNS

Pattern 1: Undue Emphasis on Significance

Words to watch: stands/serves as, is a testament/reminder, a vital/significant/crucial/pivotal/key role/moment, underscores/highlights its importance/significance, reflects broader, symbolizing its ongoing/enduring/lasting, contributing to the, setting the stage for, marking/shaping the, represents/marks a shift, key turning point, evolving landscape, focal point, indelible mark, deeply rooted

Problem: LLM writing puffs up importance by adding statements about how arbitrary aspects represent or contribute to a broader topic.

Severity: High

Before:

The Statistical Institute of Catalonia was officially established in 1989, marking a pivotal moment in the evolution of regional statistics in Spain. This initiative was part of a broader movement across Spain to decentralize administrative functions and enhance regional governance.

After:

The Statistical Institute of Catalonia was established in 1989 to collect and publish regional statistics independently from Spain's national statistics office.

Pattern 2: Undue Emphasis on Notability

Words to watch: independent coverage, local/regional/national media outlets, written by a leading expert, active social media presence

Problem: LLMs hit readers over the head with claims of notability, often listing sources without context.

Severity: Medium

Before:

Her views have been cited in The New York Times, BBC, Financial Times, and The Hindu. She maintains an active social media presence with over 500,000 followers.

After:

In a 2024 New York Times interview, she argued that AI regulation should focus on outcomes rather than methods.

Pattern 3: Superficial -ing Analyses

Words to watch: highlighting/underscoring/emphasizing..., ensuring..., reflecting/symbolizing..., contributing to..., cultivating/fostering..., encompassing..., showcasing...

Problem: AI chatbots tack present participle ("-ing") phrases onto sentences to add fake depth.

Severity: High

Before:

The temple's color palette of blue, green, and gold resonates with the region's natural beauty, symbolizing Texas bluebonnets, the Gulf of Mexico, and the diverse Texan landscapes, reflecting the community's deep connection to the land.

After:

The temple uses blue, green, and gold colors. The architect said these were chosen to reference local bluebonnets and the Gulf coast.

Pattern 4: Promotional Language

Words to watch: boasts a, vibrant, rich (figurative), profound, enhancing its, showcasing, exemplifies, commitment to, natural beauty, nestled, in the heart of, groundbreaking (figurative), renowned, breathtaking, must-visit, stunning

Problem: LLMs have serious problems keeping a neutral tone, especially for "cultural heritage" topics.

Severity: High

Before:

Nestled within the breathtaking region of Gonder in Ethiopia, Alamata Raya Kobo stands as a vibrant town with a rich cultural heritage and stunning natural beauty.

After:

Alamata Raya Kobo is a town in the Gonder region of Ethiopia, known for its weekly market and 18th-century church.

Pattern 5: Vague Attributions

Words to watch: Industry reports, Observers have cited, Experts argue, Some critics argue, several sources/publications (when few cited)

Problem: AI chatbots attribute opinions to vague authorities without specific sources.

Severity: Medium

Before:

Due to its unique characteristics, the Haolai River is of interest to researchers and conservationists. Experts believe it plays a crucial role in the regional ecosystem.

After:

The Haolai River supports several endemic fish species, according to a 2019 survey by the Chinese Academy of Sciences.

Pattern 6: Formulaic "Challenges" Sections

Words to watch: Despite its... faces several challenges..., Despite these challenges, Challenges and Legacy, Future Outlook

Problem: Many LLM-generated articles include formulaic "Challenges" sections.

Severity: Medium

Before:

Despite its industrial prosperity, Korattur faces challenges typical of urban areas, including traffic congestion and water scarcity. Despite these challenges, with its strategic location and ongoing initiatives, Korattur continues to thrive as an integral part of Chennai's growth.

After:

Traffic congestion increased after 2015 when three new IT parks opened. The municipal corporation began a stormwater drainage project in 2022 to address recurring floods.

Pattern 7: Overused AI Vocabulary

High-frequency AI words: Additionally, align with, commendable, crucial, delve, emphasizing, enduring, enhance, fostering, garner, highlight (verb), interplay, intricate/intricacies, key (adjective), landscape (abstract noun), meticulous, pivotal, showcase, tapestry (abstract noun), testament, underscore (verb), valuable, vibrant

Problem: These words appear far more frequently in post-2023 text. They often co-occur.

Severity: Medium

Before:

Additionally, a distinctive feature of Somali cuisine is the incorporation of camel meat. An enduring testament to Italian colonial influence is the widespread adoption of pasta in the local culinary landscape, showcasing how these dishes have integrated into the traditional diet.

After:

Somali cuisine also includes camel meat, which is considered a delicacy. Pasta dishes, introduced during Italian colonization, remain common, especially in the south.

Pattern 8: Copula Avoidance

Words to watch: serves as/stands as/marks/represents [a], boasts/features/offers [a]

Problem: LLMs substitute elaborate constructions for simple copulas.

Severity: Medium

Before:

Gallery 825 serves as LAAA's exhibition space for contemporary art. The gallery features four separate spaces and boasts over 3,000 square feet.

After:

Gallery 825 is LAAA's exhibition space for contemporary art. The gallery has four rooms totaling 3,000 square feet.

Pattern 9: Negative Parallelisms

Problem: Constructions like "Not only...but..." or "It's not just about..., it's..." are overused.

Severity: Low

Before:

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.

After:

The heavy beat adds to the aggressive tone.

Pattern 10: Rule of Three Overuse

Problem: LLMs force ideas into groups of three to appear comprehensive.

Severity: Low

Before:

The event features keynote sessions, panel discussions, and networking opportunities. Attendees can expect innovation, inspiration, and industry insights.

After:

The event includes talks and panels. There's also time for informal networking between sessions.

Pattern 11: Elegant Variation (Synonym Cycling)

Problem: AI has repetition-penalty code causing excessive synonym substitution.

Severity: Medium

Before:

The protagonist faces many challenges. The main character must overcome obstacles. The central figure eventually triumphs. The hero returns home.

After:

The protagonist faces many challenges but eventually triumphs and returns home.

Pattern 12: False Ranges

Problem: LLMs use "from X to Y" constructions where X and Y aren't on a meaningful scale.

Severity: Low

Before:

Our journey through the universe has taken us from the singularity of the Big Bang to the grand cosmic web, from the birth and death of stars to the enigmatic dance of dark matter.

After:

The book covers the Big Bang, star formation, and current theories about dark matter.

STYLE PATTERNS

Pattern 13: Em Dash Overuse

Problem: LLMs use em dashes (—) more than humans, mimicking "punchy" sales writing.

Severity: Low

Before:

The term is primarily promoted by Dutch institutions—not by the people themselves. You don't say "Netherlands, Europe" as an address—yet this mislabeling continues—even in official documents.

After:

The term is primarily promoted by Dutch institutions, not by the people themselves. You don't say "Netherlands, Europe" as an address, yet this mislabeling continues in official documents.

Pattern 14: Overuse of Boldface

Problem: AI chatbots emphasize phrases in boldface mechanically.

Severity: Low

Before:

It blends OKRs (Objectives and Key Results), KPIs (Key Performance Indicators), and visual strategy tools such as the Business Model Canvas (BMC) and Balanced Scorecard (BSC).

After:

It blends OKRs, KPIs, and visual strategy tools like the Business Model Canvas and Balanced Scorecard.

Pattern 15: Inline-Header Vertical Lists

Problem: AI outputs lists where items start with bolded headers followed by colons.

Severity: Low

Before:

User Experience: The user experience has been significantly improved with a new interface.
Performance: Performance has been enhanced through optimized algorithms.
Security: Security has been strengthened with end-to-end encryption.

After:

The update improves the interface, speeds up load times through optimized algorithms, and adds end-to-end encryption.

Pattern 16: Title Case in Headings

Problem: AI chatbots capitalize all main words in headings.

Severity: Low

Before:

Strategic Negotiations And Global Partnerships

After:

Strategic negotiations and global partnerships

Pattern 17: Emojis

Problem: AI chatbots often decorate headings or bullet points with emojis.

Severity: Low

Before:

🚀 Launch Phase: The product launches in Q3 💡 Key Insight: Users prefer simplicity ✅ Next Steps: Schedule follow-up meeting

After:

The product launches in Q3. User research showed a preference for simplicity. Next step: schedule a follow-up meeting.

Pattern 18: Quotation Mark Issues

Problem: AI models make two common quotation mistakes:

Using curly quotes ("...") instead of straight quotes ("...")
Using single quotes ('...') as primary delimiters in prose (from code training)

Severity: Low

Before:

He said "the project is on track" but others disagreed. She stated, 'This is the final version.'

After:

He said "the project is on track" but others disagreed. She stated, "This is the final version."

COMMUNICATION PATTERNS

Pattern 19: Collaborative Communication Artifacts

Words to watch: I hope this helps, Of course!, Certainly!, You're absolutely right!, Would you like..., let me know, here is a...

Problem: Text meant as chatbot correspondence gets pasted as content.

Severity: Critical

Before:

Here is an overview of the French Revolution. I hope this helps! Let me know if you'd like me to expand on any section.

After:

The French Revolution began in 1789 when financial crisis and food shortages led to widespread unrest.

Pattern 20: Knowledge-Cutoff Disclaimers

Words to watch: as of [date], Up to my last training update, While specific details are limited/scarce..., based on available information...

Problem: AI disclaimers about incomplete information get left in text.

Severity: Critical

Before:

While specific details about the company's founding are not extensively documented in readily available sources, it appears to have been established sometime in the 1990s.

After:

The company was founded in 1994, according to its registration documents.

Pattern 21: Sycophantic Tone

Problem: Overly positive, people-pleasing language.

Severity: Critical

Before:

Great question! You're absolutely right that this is a complex topic. That's an excellent point about the economic factors.

After:

The economic factors you mentioned are relevant here.

FILLER AND HEDGING

Pattern 22: Filler Phrases

Problem: Wordy constructions that add no value.

Severity: Low

Before → After:

"In order to achieve this goal" → "To achieve this"
"Due to the fact that it was raining" → "Because it was raining"
"At this point in time" → "Now"
"In the event that you need help" → "If you need help"
"The system has the ability to process" → "The system can process"
"It is important to note that the data shows" → "The data shows"

Pattern 23: Excessive Hedging

Problem: Over-qualifying statements.

Severity: Low

Before:

It could potentially possibly be argued that the policy might have some effect on outcomes.

After:

The policy may affect outcomes.

Pattern 24: Generic Positive Conclusions

Problem: Vague upbeat endings.

Severity: Low

Before:

The future looks bright for the company. Exciting times lie ahead as they continue their journey toward excellence. This represents a major step in the right direction.

After:

The company plans to open two more locations next year.

Pattern 25: AI Signatures in Code

Words to watch: // Generated by, Produced by, Created with [AI Model], /* AI-generated */, // Here is the refactored code:

Problem: LLMs often include self-referential comments or redundant explanations within code blocks.

Severity: Critical

Before:

// Generated by ChatGPT
// This function adds two numbers
function add(a, b) {
  return a + b;
}

After:

function add(a, b) {
  return a + b;
}

Pattern 26: Over-Structuring

Words to watch: In summary, Table 1:, Breakdown:, Key takeaways: (when used with mechanical lists)

Problem: AI-generated text often uses rigid, non-human formatting to present simple information that a human would describe narratively.

Severity: Low

Before:

Performance Comparison:

Speed: High

Stability: Excellent

Memory: Low

After:

The system is fast and stable with low memory overhead.

Pattern 27: Technical Literal Preservation

Rule: Never modify the following, even if they match AI patterns:

Anything inside inline code/backticks (e.g., foo_bar, --flag, path/to/file)
Anything inside fenced code blocks
URLs (including query strings), file paths, version strings, hashes/IDs
API names, identifiers, CLI commands/flags, config keys, error messages

Severity: Critical (must preserve)

Example:

The --verbose flag enables detailed logging. See docs/api.md for more.

Do NOT change to:

The verbose option enables detailed logging. See the API documentation for more.

Pattern 28: Persuasive Tropes

Words to watch: The real question is, At its core, What this really means is, The truth is

Problem: Frames ordinary claims as revelations. The sentence after these phrases almost always restates something already said.

Severity: Low

Before:

The real question is whether this approach will work. At its core, this is about making better decisions.

After:

This approach will work if we implement it correctly. This is about making better decisions.

Not a problem when: Used in legitimate contexts like op-eds or presentation scripts.

Pattern 29: Signposting

Words to watch: Let's dive in, Here's what you need to know, Let's explore, In this article we'll

Problem: The model announces what it's about to do instead of doing it.

Severity: Low

Before:

Let's dive in and explore the key features. Here's what you need to know about the system.

After:

The system has three key features: speed, reliability, and security.

Not a problem when: Used in legitimate contexts like presentation scripts or tutorials.

Pattern 30: Fragmented Headers

Problem: A short generic sentence appears right after a heading (e.g., "Speed matters.") before the actual paragraph. Adds nothing the heading doesn't already say.

Severity: Low

Before:

## Performance

Speed matters. The system processes requests in under 100ms.

After:

## Performance

The system processes requests in under 100ms.

Not a problem when: Used in legitimate contexts like op-eds or persuasive writing.

SEVERITY CLASSIFICATION

Critical (immediate AI detection)

Pattern 19: Collaborative communication artifacts
Pattern 20: Knowledge-cutoff disclaimers
Pattern 21: Sycophantic tone
Pattern 25: AI signatures in code
Pattern 27: Technical literal preservation (must preserve)

High (strong AI signals)

Pattern 1: Undue emphasis on significance
Pattern 3: Superficial -ing analyses
Pattern 4: Promotional language
Pattern 31: Extended thinking tags

Medium (moderate AI signals)

Pattern 2: Undue emphasis on notability
Pattern 5: Vague attributions
Pattern 6: Formulaic "Challenges" sections
Pattern 7: Overused AI vocabulary
Pattern 8: Copula avoidance
Pattern 11: Elegant variation
Pattern 32: JSON mode artifacts
Pattern 33: Tool use documentation

Low (weak AI signals)

Pattern 9: Negative parallelisms
Pattern 10: Rule of three overuse
Pattern 12: False ranges
Pattern 13: Em dash overuse
Pattern 14: Overuse of boldface
Pattern 15: Inline-header lists
Pattern 16: Title case in headings
Pattern 17: Emojis
Pattern 18: Quotation mark issues
Pattern 22: Filler phrases
Pattern 23: Excessive hedging
Pattern 24: Generic positive conclusions
Pattern 26: Over-structuring
Pattern 28: Persuasive tropes
Pattern 29: Signposting
Pattern 30: Fragmented headers
Pattern 34: Over-polished conclusions
Pattern 14: Overuse of boldface
Pattern 15: Inline-header lists
Pattern 16: Title case in headings
Pattern 17: Emojis
Pattern 18: Quotation mark issues
Pattern 22: Filler phrases
Pattern 23: Excessive hedging
Pattern 24: Generic positive conclusions
Pattern 26: Over-structuring
Pattern 28: Persuasive tropes
Pattern 29: Signposting
Pattern 30: Fragmented headers

Module Version: 3.1.0 Last Updated: 2026-04-04 Patterns: 34 (30 original + 4 new for 2025-2026 LLM variants) Source: Wikipedia "Signs of AI writing" + Humanizer community contributions + 2025-2026 LLM analysis

Pattern 31: Extended Thinking Tags

Problem: Modern reasoning models (GPT-4.5, Claude 4, DeepSeek) produce visible <thinking>, <reflection>, or <analysis> tags in output. These are model introspection artifacts, not user content.

Severity: High

Words to watch: <thinking>, </thinking>, <reflection>, <reflection>, <analysis>, </analysis>, <reasoning>, </reasoning>

Before:

The solution involves... <thinking>I need to consider the edge cases...</thinking> Let me explain.

After:

The solution involves several key factors. Let me explain.

Not a problem when: Model is explicitly asked to show its reasoning in structured format.

Pattern 32: JSON Mode Artifacts

Problem: Models forced into JSON output often produce overly structured responses with explicit JSON schema comments, "Here is the JSON" preambles, or unnecessary escaping.

Severity: Medium

Words to watch: "Here is the", "```json", "```", "JSON:", "as requested"

Before:

Here's the JSON as you requested:
{ "name": "example", "value": 123 }

After:

{ "name": "example", "value": 123 }

Not a problem when: Actual API responses or configuration files.

Pattern 33: Tool Use Documentation

Problem: Models with tool-use capabilities add verbose "I will use tool X" preambles before executing actions, especially in agentic workflows.

Severity: Medium

Words to watch: "I will use", "I am going to use", "Calling function", "Executing", "Running", "invoking"

Before:

I will use the file read tool to access the configuration. Let me read the file now.

After:

Reading the configuration file now.

Not a problem when: Explicit tutorials or documentation about tool usage.

Pattern 34: Over-Polished Conclusions

Problem: Newer models produce excessively diplomatic conclusions with "hope this helps", "please let me know if you need anything else", and other service-industry language that feels inhuman.

Severity: Low

Words to watch: "hope this helps", "let me know if", "happy to help", "feel free to", "don't hesitate to", "anytime"

Before:

That should solve your issue! Let me know if you run into anything else. Happy to help!

After:

That should solve your issue.

Not a problem when: Genuine customer service contexts.

TECHNICAL MODULE

module_id: technical version: 3.0.0 description: Technical module for code and engineering documentation applies_to: code, technical docs, API docs, READMEs severity_levels:

Critical
High
Medium
Low

Module: Technical

Description

This module applies to code, technical documentation, API documentation, READMEs, and engineering writing. It preserves technical precision while removing AI voice patterns.

When to Apply:

Code comments and docstrings
API documentation
README files
Technical specifications
Engineering design docs
Commit messages
Code review comments

When NOT to Apply:

Creative writing
Marketing copy
Personal blogs (unless technical)

TECHNICAL NUANCE

Expertise isn't slop. In technical contexts, words like "crucial", "pivotal", or "critical" are sometimes the exact right terms for describing requirements, dependencies, or system behavior.

Rule: If a word is required for technical accuracy, keep it. If it's there to add fake "gravitas" or marketing polish, cut it.

Examples

Acceptable (technical precision):

"The authentication step is critical - without it, any user can access admin endpoints."

Unacceptable (fake gravitas):

"The authentication step is absolutely critical and plays a pivotal role in the ever-evolving landscape of modern security."

TECHNICAL LITERAL PRESERVATION

CRITICAL RULE: Never modify the following, even if they match AI patterns:

1. Code and Identifiers

Anything inside inline code/backticks: `foo_bar`, `--flag`, `path/to/file`
Anything inside fenced code blocks (```)
Function names, class names, variable names
API endpoints, HTTP methods
CLI commands and flags
Configuration keys
Error messages (exact text)
Stack traces

Before (AI-generated comment):

// This function adds two numbers together and returns the sum
function add(a, b) {
  return a + b;
}

After (concise, human):

// Adds two numbers
function add(a, b) {
  return a + b;
}

NOT (over-correcting):

// do math
function add(a, b) {
  return a + b;
}

2. Technical Specifications

Version strings: v2.3.0, Node 20.x, Python 3.11
URLs and query strings
File paths (absolute and relative)
Hashes, IDs, UUIDs
Database schemas, table names
Protocol names: HTTP/2, WebSocket, gRPC

3. Technical Accuracy

Mathematical formulas
Algorithm descriptions
Data structures
Type signatures
Interface definitions

CODE COMMENT PATTERNS

Pattern T1: Redundant Comment Explanations

Problem: AI generates comments that explain obvious code, adding noise without value.

Severity: Low

Before:

// Initialize the counter variable to zero
let counter = 0;

// Loop through the array and process each item
for (let i = 0; i < items.length; i++) {
  // Process the current item
  processItem(items[i]);
}

After:

let counter = 0;

for (let i = 0; i < items.length; i++) {
  processItem(items[i]);
}

Pattern T2: AI Signatures in Code

Problem: LLMs include self-referential comments or generation markers.

Severity: Critical

Words to watch:

// Generated by
// Created with
/* AI-generated */
// This code was written by
// Here is the refactored code:

Before:

// Generated by GitHub Copilot
// This function validates user input
function validateInput(input) {
  // Check if input is valid
  return input !== null;
}

After:

function validateInput(input) {
  return input !== null;
}

Pattern T3: Over-Explained Docstrings

Problem: AI generates verbose docstrings that restate the obvious.

Severity: Low

Before:

def calculate_total(items):
    """
    This function calculates the total sum of all items in the list.

    Args:
        items: A list of numeric items to sum up

    Returns:
        The total sum of all items as a number
    """
    return sum(items)

After:

def calculate_total(items):
    """Sum of items."""
    return sum(items)

Pattern T4: Filler in Technical Writing

Problem: AI adds unnecessary hedging and filler to technical docs.

Severity: Low

Before:

"It's important to note that this function should be called before initializing the database connection. Please be aware that failure to do so may potentially result in connection errors."

After:

"Call this function before initializing the database connection. Failure to do so will result in connection errors."

API DOCUMENTATION PATTERNS

Pattern T5: Promotional API Descriptions

Problem: AI uses marketing language in API docs instead of clear technical descriptions.

Severity: Medium

Words to watch:

"powerful", "robust", "seamless", "effortless"
"game-changing", "revolutionary", "cutting-edge"
"unlock the full potential", "take advantage of"

Before:

"Our powerful API provides seamless integration with your existing systems, unlocking the full potential of your data pipeline."

After:

"The API integrates with existing systems via REST endpoints. See /api/v1/pipeline for data ingestion."

Pattern T6: Vague Technical Descriptions

Problem: AI uses vague language instead of specific technical details.

Severity: Medium

Before:

"This endpoint handles various types of data processing operations efficiently."

After:

"POST /api/v1/process accepts JSON payloads up to 10MB and returns processed results within 500ms (p95)."

README PATTERNS

Pattern T7: Generic Positive Conclusions

Problem: READMEs end with vague upbeat statements instead of actionable next steps.

Severity: Low

Before:

"We're excited to see what you'll build with this tool! The future looks bright as we continue to improve and add features. Happy coding!"

After:

"See examples/ for usage examples. Report issues on the GitHub tracker."

Pattern T8: Over-Structured Installation Sections

Problem: AI uses rigid formatting for simple installation steps.

Severity: Low

Before:

Installation Steps

Prerequisites: Ensure Node.js is installed

Clone Repository: Run git clone

Install Dependencies: Run npm install

Verify Installation: Run npm test

After:

Installation

Requires Node.js 20+.
git clone <repo>
npm install
npm test  # verify installation

COMMIT MESSAGE PATTERNS

Pattern T9: Vague Commit Messages

Problem: AI generates generic commit messages that don't explain the change.

Severity: Medium

Before:

"Update code to improve performance and fix issues"

After:

"perf: reduce database queries in user lookup by 40%"

Pattern T10: Over-Explained Commit Bodies

Problem: AI adds unnecessary context or apologies in commit messages.

Severity: Low

Before:

"This commit fixes the bug that was causing issues. I hope this resolves the problem. Let me know if there are any other issues."

After:

"Fix: null pointer in user service when email is missing"

CODE REVIEW PATTERNS

Pattern T11: Sycophantic Review Comments

Problem: AI uses overly polite or apologetic language in code reviews.

Severity: Low

Before:

"Great work on this! I really like your approach. Just a small suggestion - would you mind considering adding a test for this edge case? No pressure though!"

After:

"Consider adding a test for the null input edge case."

Pattern T12: Hedged Technical Feedback

Problem: AI hedges technical feedback unnecessarily.

Severity: Low

Before:

"I'm not entirely sure, but it seems like this might potentially cause a memory leak if the listener isn't cleaned up properly."

After:

"This will cause a memory leak if the listener isn't cleaned up. Add removeEventListener in the cleanup function."

ERROR HANDLING PATTERNS

Pattern T13: Vague Error Messages

Problem: AI generates generic error messages that don't help debugging.

Severity: Medium

Before:

"An error occurred while processing your request. Please try again later."

After:

"Database connection failed: timeout after 30s. Check network connectivity and retry."

Pattern T14: Over-Apologetic Errors

Problem: AI adds unnecessary apologies to error messages.

Severity: Low

Before:

"We're sorry, but unfortunately an unexpected error has occurred. We apologize for the inconvenience."

After:

"Internal error: failed to parse JSON at line 42, column 15."

SEVERITY CLASSIFICATION

Critical (must fix)

Pattern T2: AI signatures in code

High (strong AI signals)

None for technical module

Medium (moderate AI signals)

Pattern T5: Promotional API descriptions
Pattern T6: Vague technical descriptions
Pattern T13: Vague error messages

Low (weak AI signals)

Pattern T1: Redundant comment explanations
Pattern T3: Over-explained docstrings
Pattern T4: Filler in technical writing
Pattern T7: Generic positive conclusions
Pattern T8: Over-structured installation
Pattern T9: Vague commit messages
Pattern T10: Over-explained commit bodies
Pattern T11: Sycophantic review comments
Pattern T12: Hedged technical feedback
Pattern T14: Over-apologetic errors

TECHNICAL WRITING BEST PRACTICES

Do

Use simple, direct language
Provide specific technical details
Include working code examples
Document edge cases and error conditions
Use consistent terminology
Link to related documentation

Don't

Use marketing language in technical docs
Add unnecessary hedging or apologies
Explain obvious code
Use vague descriptions ("various", "many", "several")
Include AI signatures or generation markers

Module Version: 3.0.0 Last Updated: 2026-03-03 Applies to: Code, technical docs, API docs, READMEs, commit messages, code reviews

ACADEMIC MODULE

module_id: academic version: 3.0.0 description: Academic module for papers, essays, and formal research prose applies_to: research papers, essays, dissertations, grant proposals severity_levels:

Critical
High
Medium
Low

Module: Academic

Description

This module applies to academic writing: research papers, essays, dissertations, grant proposals, and formal research prose. It maintains scholarly rigor while removing AI voice patterns.

When to Apply:

Research papers
Academic essays
Dissertations and theses
Grant proposals
Literature reviews
Conference submissions

When NOT to Apply:

Creative writing
Technical documentation
Business communications

ACADEMIC VOICE

Scholarly precision matters. Academic writing has specific conventions: hedging where appropriate, acknowledging limitations, citing sources properly. The goal is to remove AI patterns while preserving legitimate academic style.

Rule: Keep legitimate academic hedging ("may suggest", "appears to indicate"). Remove AI filler ("it is worth noting that", "it is important to emphasize").

ACADEMIC PATTERNS

Pattern A1: Vague Literature Citations

Problem: AI attributes claims to vague authorities without specific citations.

Severity: High

Words to watch:

"Studies have shown"
"Research indicates"
"Experts agree"
"It has been demonstrated"

Before:

Studies have shown that climate change significantly impacts biodiversity. Research indicates that immediate action is necessary.

After:

Smith et al. (2023) found that climate change reduced local biodiversity by 40% over two decades. Immediate conservation measures are recommended (Jones, 2024).

Pattern A2: Formulaic Literature Review Sections

Problem: AI generates rigid, template-like literature review paragraphs.

Severity: Medium

Before:

Previous Research: Previous research has explored this topic extensively. Current Gap: However, current research has limitations. Our Contribution: Our study addresses these gaps.

After:

Prior work established the foundation for this study (Smith, 2022; Jones, 2023). However, these studies were limited to laboratory conditions. Our field study addresses this limitation.

Pattern A3: Over-Hedging

Problem: AI over-qualifies statements beyond legitimate academic caution.

Severity: Low

Before:

It could potentially be suggested that the results may possibly indicate a trend that might warrant further investigation.

After:

The results suggest a trend warranting further investigation.

Pattern A4: Generic Conclusions

Problem: AI ends papers with vague statements about "future research" and "broader implications."

Severity: Medium

Before:

In conclusion, this study has provided valuable insights. Future research should explore these findings further. The implications are significant for the field.

After:

This study demonstrates X under conditions Y. Future work should test whether X holds in real-world settings. The methodology may apply to similar problems in Z domain.

Pattern A5: Promotional Abstract Language

Problem: AI uses marketing language in abstracts instead of clear findings.

Severity: Medium

Words to watch:

"groundbreaking", "novel", "innovative"
"comprehensive", "extensive", "thorough"
"significant contributions", "valuable insights"

Before:

This groundbreaking study provides comprehensive insights into the novel methodology, making significant contributions to the field.

After:

We present a method achieving 95% accuracy on dataset X, improving on prior work by 12%.

Pattern A6: Filler in Methodology

Problem: AI adds unnecessary words to methodology descriptions.

Severity: Low

Before:

In order to achieve the goal of analyzing the data, we employed the use of statistical methods.

After:

We analyzed the data using ANOVA.

Pattern A7: Artificial Signposting

Problem: AI uses excessive structural markers in academic writing.

Severity: Low

Words to watch:

"Firstly", "Secondly", "Thirdly"
"In the first section", "In the second section"
"This paper is organized as follows"

Before:

Firstly, we review the literature. Secondly, we describe our methodology. Thirdly, we present results.

After:

We review the literature (Section 2), describe our methodology (Section 3), and present results (Section 4).

Pattern A8: Vague Quantitative Claims

Problem: AI makes imprecise quantitative statements.

Severity: Medium

Before:

A significant number of participants showed improvement.

After:

73 of 100 participants (73%) showed improvement (p < 0.01).

CITATION AND REFERENCING

Pattern A9: Fake or Inaccurate Citations

Problem: AI generates plausible-looking but fake or inaccurate citations.

Severity: Critical

Action: Verify every citation against real databases (Google Scholar, DOI, PubMed).

Before:

(Smith et al., 2023) found significant effects.

After:

[Verify: Does Smith et al. 2023 actually exist? Check DOI.]

Pattern A10: Citation Padding

Problem: AI adds unnecessary citations to appear authoritative.

Severity: Low

Before:

Climate change is a serious problem [1-15].

After:

Global average temperature has increased 1.1°C since 1880 (NASA, 2023).

SEVERITY CLASSIFICATION

Critical (must fix)

Pattern A9: Fake or inaccurate citations

High (strong AI signals)

Pattern A1: Vague literature citations

Medium (moderate AI signals)

Pattern A2: Formulaic literature review sections
Pattern A4: Generic conclusions
Pattern A5: Promotional abstract language
Pattern A8: Vague quantitative claims

Low (weak AI signals)

Pattern A3: Over-hedging
Pattern A6: Filler in methodology
Pattern A7: Artificial signposting
Pattern A10: Citation padding

ACADEMIC WRITING BEST PRACTICES

Do

Cite specific sources with verifiable references
Use appropriate hedging for claims
Report exact statistics and p-values
Acknowledge limitations clearly
Use field-standard terminology
Follow journal/conference style guides

Don't

Use vague citations ("studies have shown")
Add promotional language ("groundbreaking", "novel")
Over-hedge beyond legitimate academic caution
Pad citations unnecessarily
Use marketing language in abstracts

Module Version: 3.0.0 Last Updated: 2026-03-03 Applies to: Research papers, essays, dissertations, grant proposals, literature reviews

GOVERNANCE MODULE

module_id: governance version: 3.0.0 description: Governance module for policy, risk, and compliance writing applies_to: policies, risk assessments, compliance docs, legal writing severity_levels:

Critical
High
Medium
Low

Module: Governance

Description

This module applies to governance writing: policies, risk assessments, compliance documentation, legal writing, and regulatory submissions. It maintains precision and formality while removing AI voice patterns.

When to Apply:

Company policies
Risk assessments
Compliance documentation
Legal contracts
Regulatory submissions
Board reports

When NOT to Apply:

Creative writing
Marketing materials
Informal communications

GOVERNANCE VOICE

Precision and clarity are critical. Governance documents have legal and regulatory implications. Remove AI patterns while preserving necessary formality and precision.

Rule: Keep required legal/formal language. Remove AI filler, vague attributions, and promotional phrasing.

GOVERNANCE PATTERNS

Pattern G1: Vague Policy Language

Problem: AI uses imprecise language in policies where specificity is required.

Severity: High

Before:

Employees should generally endeavor to maintain appropriate security practices where feasible.

After:

Employees must enable two-factor authentication on all company accounts.

Pattern G2: Hedged Risk Statements

Problem: AI over-hedges risk statements, weakening accountability.

Severity: High

Before:

There may potentially be some risk that data could possibly be compromised in certain circumstances.

After:

Risk: Unencrypted data in transit may be intercepted. Likelihood: Medium. Impact: High.

Pattern G3: Promotional Compliance Language

Problem: AI uses marketing language in compliance documents.

Severity: Medium

Words to watch:

"commitment to excellence", "dedication to"
"best-in-class", "industry-leading"
"unwavering commitment", "paramount importance"

Before:

Our unwavering commitment to data protection demonstrates our dedication to best-in-class security practices.

After:

We comply with GDPR Article 32 (security of processing) through encryption, access controls, and regular audits.

Pattern G4: Vague Attributions in Policy

Problem: AI attributes requirements to vague authorities.

Severity: High

Words to watch:

"Industry standards require"
"Regulations state"
"Experts recommend"

Before:

Industry standards require regular security assessments.

After:

SOC 2 Type II requires annual security assessments (AICPA, 2023).

Pattern G5: Formulaic "Future Outlook" Sections

Problem: AI adds generic forward-looking statements to governance docs.

Severity: Low

Before:

Looking ahead, we remain committed to continuous improvement. The future looks bright as we enhance our governance framework.

After:

Next review date: 2027-03-03. Responsible: Chief Compliance Officer.

Pattern G6: Over-Structured Risk Matrices

Problem: AI uses rigid formatting for risk descriptions that humans would write narratively.

Severity: Low

Before:

Risk Category: Cybersecurity Likelihood: High Impact: Critical Mitigation: Implement controls

After:

Cybersecurity risk is high with critical potential impact. Mitigation: implement access controls, encryption, and monitoring.

Pattern G7: Filler in Legal Writing

Problem: AI adds unnecessary words to legal/policy text.

Severity: Medium

Before:

In the event that an employee fails to comply with the provisions set forth herein, disciplinary action may be taken.

After:

Employees who violate this policy face disciplinary action, up to and including termination.

Pattern G8: Generic Positive Conclusions

Problem: AI ends governance docs with vague upbeat statements.

Severity: Low

Before:

We are confident that these measures will ensure a bright and secure future for our organization.

After:

This policy takes effect 2026-04-01. Questions: compliance@company.com

COMPLIANCE AND REGULATORY

Pattern G9: Vague Regulatory References

Problem: AI references regulations without specific sections or requirements.

Severity: High

Before:

We comply with all applicable data protection regulations.

After:

We comply with:

GDPR (EU) 2016/679: Articles 5, 6, 32

CCPA (California): Section 1798.100

HIPAA (US): 45 CFR Part 160

Pattern G10: Missing Accountability

Problem: AI policies lack clear ownership and enforcement.

Severity: Medium

Before:

This policy should be followed by all employees.

After:

Owner: Chief Compliance Officer Applies to: All employees, contractors, vendors Enforcement: HR and Legal Violations: Report to compliance@company.com

SEVERITY CLASSIFICATION

Critical (must fix)

None for governance module (precision varies by context)

High (strong AI signals)

Pattern G1: Vague policy language
Pattern G2: Hedged risk statements
Pattern G4: Vague attributions in policy
Pattern G9: Vague regulatory references

Medium (moderate AI signals)

Pattern G3: Promotional compliance language
Pattern G7: Filler in legal writing
Pattern G10: Missing accountability

Low (weak AI signals)

Pattern G5: Formulaic "Future Outlook" sections
Pattern G6: Over-structured risk matrices
Pattern G8: Generic positive conclusions

GOVERNANCE WRITING BEST PRACTICES

Do

Use precise, unambiguous language
Cite specific regulations and sections
Define clear ownership and accountability
Include enforcement mechanisms
Use consistent terminology
Set specific review dates

Don't

Use vague language ("should", "may", "where feasible")
Add promotional phrasing
Hedge risk statements unnecessarily
Reference vague authorities
End with generic positive conclusions

Module Version: 3.0.0 Last Updated: 2026-03-03 Applies to: Policies, risk assessments, compliance docs, legal writing, regulatory submissions

REASONING MODULE

module_id: reasoning version: 1.0.0 description: Module for detecting and addressing LLM reasoning failures severity_levels:

Critical
High
Medium
Low

Humanizer Reasoning Module: LLM Reasoning Failures

This module identifies and addresses reasoning failures in Large Language Model (LLM) outputs that manifest as detectable patterns in the generated text.

Description

Reasoning Failure Categories

Pattern R1: Depth-Dependent Reasoning Failures

Problem: Accuracy decreases as reasoning chain lengthens
Severity: Medium
Action: Simplify complex explanations, remove tangential content, ensure focus

Pattern R2: Context-Switching Failures

Problem: Difficulty maintaining coherence across different knowledge domains
Severity: High
Action: Smooth transitions between topics, maintain consistent register and tone

Pattern R3: Temporal Reasoning Limitations

Problem: Errors in temporal sequence or causal reasoning tasks
Severity: Medium
Action: Clarify chronological order, strengthen causal connections

Pattern R4: Abstraction-Level Mismatches

Problem: Difficulty maintaining appropriate level of abstraction
Severity: Medium
Action: Bridge abstraction gaps with clear connections

5. Logical Fallacy Susceptibility

Sign: Systematic reasoning errors that contradict formal logic
Action: Identify and correct logical inconsistencies

6. Quantitative Reasoning Deficits

Sign: Errors in numerical computation or quantitative understanding
Action: Flag questionable numerical claims for review

7. Self-Consistency Failures

Sign: Internal contradictions within a single output
Action: Identify and resolve internal contradictions

8. Verification and Checking Deficiencies

Sign: Lack of internal verification mechanisms
Action: Add appropriate qualifiers, acknowledge uncertainties

APPLICATION RULES

When to Apply

When text quality critically depends on logical consistency
When dealing with technical, academic, or factual content
When surface-level fixes are insufficient for naturalness

When Not to Apply

For general casual writing where logical depth isn't critical
When computational efficiency is paramount
When the text is already logically sound

INTEGRATION WITH OTHER MODULES

Core Humanizer addresses surface-level writing quality issues
Reasoning module addresses deeper logical consistency issues
Both modules can operate independently or in combination
Reasoning module defers to Core for surface-level fixes

QUALITY STANDARDS

All reasoning diagnostics must meet these standards:

Corrections must be logically sound
Claims must be verifiable or appropriately qualified
Changes must improve accuracy and consistency
Evidence for diagnoses must be documented

FilesExpand file tree

SKILL_PROFESSIONAL.md

Latest commit

History

SKILL_PROFESSIONAL.md

File metadata and controls

Humanizer: Remove AI Writing Patterns

Humanizer Pro: Professional Editing

Modules

ROUTING LOGIC

Professional Tone

Your Task

CLARITY AND TONE

Signs the writing is still flat

What to aim for

Technical Nuance

CORE PATTERNS MODULE

Module: Core Patterns

Description

Your Task

PERSONALITY AND SOUL

Signs of soulless writing (even if technically "clean")

How to add voice

Before (clean but soulless)

After (has a pulse)

CONTENT PATTERNS

Pattern 1: Undue Emphasis on Significance

Pattern 2: Undue Emphasis on Notability

Pattern 3: Superficial -ing Analyses

Pattern 4: Promotional Language

Pattern 5: Vague Attributions

Pattern 6: Formulaic "Challenges" Sections

Pattern 7: Overused AI Vocabulary

Pattern 8: Copula Avoidance

Pattern 9: Negative Parallelisms

Pattern 10: Rule of Three Overuse

Pattern 11: Elegant Variation (Synonym Cycling)

Pattern 12: False Ranges

STYLE PATTERNS

Pattern 13: Em Dash Overuse

Pattern 14: Overuse of Boldface

Pattern 15: Inline-Header Vertical Lists

Pattern 16: Title Case in Headings

Strategic Negotiations And Global Partnerships

Strategic negotiations and global partnerships

Pattern 17: Emojis

Pattern 18: Quotation Mark Issues

COMMUNICATION PATTERNS

Pattern 19: Collaborative Communication Artifacts

Pattern 20: Knowledge-Cutoff Disclaimers

Pattern 21: Sycophantic Tone

FILLER AND HEDGING

Pattern 22: Filler Phrases

Pattern 23: Excessive Hedging

Pattern 24: Generic Positive Conclusions

Pattern 25: AI Signatures in Code

Pattern 26: Over-Structuring

Pattern 27: Technical Literal Preservation

Pattern 28: Persuasive Tropes

Pattern 29: Signposting

Pattern 30: Fragmented Headers

SEVERITY CLASSIFICATION

Critical (immediate AI detection)

High (strong AI signals)

Medium (moderate AI signals)

Low (weak AI signals)

Pattern 31: Extended Thinking Tags

Pattern 32: JSON Mode Artifacts

Pattern 33: Tool Use Documentation

Pattern 34: Over-Polished Conclusions

TECHNICAL MODULE

Module: Technical

Description

TECHNICAL NUANCE

Examples

TECHNICAL LITERAL PRESERVATION

1. Code and Identifiers

2. Technical Specifications

3. Technical Accuracy

CODE COMMENT PATTERNS