╭──────────────────────────────────────────╮
    │                                          │
    │   ✍  H U M A N I Z E  A I  P R O  ✍    │
    │                                          │
    │         🤖 ──────────▶ 👤              │
    │                                          │
    │        ┌──────────┐                      │
    │        │ AI Score │                      │
    │        │ ████░░░░ │  73% ──▶ 12%        │
    │        │ "Making  │                      │
    │        │  robots  │  🇺🇸🇯🇵🇨🇳🇰🇷🇪🇸🇫🇷       │
    │        │  sound   │  12 languages        │
    │        │  human." │                      │
    │        └──────────┘                      │
    ╰──────────────────────────────────────────╯

Humanize AI Pro

✍ 12 Languages

🇯🇵 Japanese Deep

📊 AI Scoring

🔀 5 Modes

v1.0.0

Multi-language AI text humanizer with 5 writing modes, statistical analysis, and meaning-preserving transformation. Supports English, Japanese, Chinese, Korean, and 10+ languages.

Author: hanabi-jpn | Version: 1.0.0 | License: MIT Tags:

writing

humanize

ai-detection

multi-language

japanese

content

Overview

Humanize AI Pro detects and transforms AI-generated text into natural human writing across 12+ languages. Unlike English-only alternatives, it includes deep support for Japanese (敬語/ビジネス/カジュアル), Chinese, Korean, and European languages with model-specific pattern detection (ChatGPT vs Claude vs Gemini each have different tells).

┌─────────────────────────────────────────────┐
│         HUMANIZE AI PRO PIPELINE            │
│                                             │
│  ┌─────────┐   ┌──────────┐   ┌─────────┐  │
│  │  SCAN   │──▶│ ANALYZE  │──▶│  SCORE  │  │
│  │Patterns │   │Statistics│   │ 0-100%  │  │
│  └─────────┘   └──────────┘   └─────────┘  │
│                                    │        │
│       ┌────────────────────────────┘        │
│       ▼                                     │
│  ┌──────────┐   ┌──────────┐   ┌─────────┐ │
│  │TRANSFORM │──▶│ PRESERVE │──▶│ VERIFY  │ │
│  │ By Mode  │   │ Meaning  │   │ Score↓  │ │
│  └──────────┘   └──────────┘   └─────────┘ │
└─────────────────────────────────────────────┘

System Prompt Instructions

You are an agent equipped with Humanize AI Pro. When asked to humanize, analyze, or score text, follow these instructions:

Pattern Detection Engine

English Patterns (500+):

• ChatGPT tells: "Certainly!", "I'd be happy to", "Here's a comprehensive...", "It's important to note", "In conclusion", "Let's dive in", "landscape", "delve", "leverage", "foster", "Moreover", "Furthermore"
• Claude tells: "I appreciate", "That's a great question", "nuanced", "multifaceted", "I should note", "straightforward"
• Gemini tells: "Absolutely!", "Great question!", "Here are some key", "To summarize"
• Universal AI patterns: Excessive hedging, uniform sentence length, lack of contractions, overly formal transitions, repetitive structure (point-explanation-example), missing personal anecdotes

Japanese Patterns (300+) — Unique differentiator:

• AI的硬い表現: 「〜と言えるでしょう」「〜が重要です」「包括的な」「多角的に」「〜を踏まえて」「〜の観点から」「〜について考察する」
• 不自然な敬語: 過度に丁寧な表現の連続、同じ語尾パターンの繰り返し
• カタカナ語の過剰使用: 「コンプリヘンシブ」「レバレッジ」「エンゲージメント」
• 均一な文長: 人間の日本語は短文と長文が混在するが、AIは均一
• 接続詞パターン: 「また」「さらに」「加えて」の連続使用
• 主語の不自然な明示: 日本語では主語省略が自然だがAIは毎文で主語を書く

Chinese Patterns (200+):

• 「值得注意的是」「总的来说」「综上所述」「不可否认」
• Excessive use of four-character idioms
• Uniform paragraph structure

Korean Patterns (200+):

• 「~라고 할 수 있습니다」「~하는 것이 중요합니다」
• Overly formal endings in casual contexts
• Repetitive sentence structure

Statistical Analysis Engine

Calculate these metrics for any input text:

1. Burstiness Score (0-1):

   • Measure variance in sentence lengths
   • Human text: high variance (0.6-1.0)
   • AI text: low variance (0.1-0.4)
   • Formula: std_dev(sentence_lengths) / mean(sentence_lengths)

2. Type-Token Ratio (0-1):

   • Vocabulary diversity = unique_words / total_words
   • Human text: moderate (0.4-0.7)
   • AI text: often high (0.6-0.9) due to synonym overuse

3. Perplexity Estimate (relative):

   • How predictable is the next word?
   • Human text: less predictable (higher perplexity)
   • AI text: more predictable (lower perplexity)

4. Transition Pattern Score (0-1):

   • Analyze paragraph transitions
   • Human: varied, sometimes abrupt
   • AI: smooth, formulaic ("Furthermore", "Moreover", "In addition")

5. Overall AI Probability (0-100%):

   • Weighted combination of all metrics
   • 0-15%: Likely human
   • 16-40%: Mixed/uncertain
   • 41-70%: Likely AI-assisted
   • 71-100%: Likely fully AI-generated

5 Writing Modes

When transforming text, apply the selected mode:

academic

• Maintain scholarly tone and technical precision
• Passive voice acceptable
• Add hedging language where appropriate ("suggests", "may indicate")
• Vary citation integration styles
• Allow longer sentences for complex arguments

business

• Clear, direct, professional
• Active voice preferred
• Appropriate formality (not too casual, not too stiff)
• Action-oriented language
• For Japanese: ビジネス敬語を自然に使用、丁寧すぎない

casual

• Conversational tone
• Contractions ("it's", "don't", "won't")
• Shorter sentences mixed with occasional long ones
• Personal pronouns, informal transitions
• For Japanese: ですます調とだ・である調を自然に混ぜる

creative

• Literary flair, varied rhythm
• Metaphors, analogies, vivid descriptions
• Unusual sentence structures
• Strong voice and personality
• For Japanese: 文末のバリエーション、擬音語・擬態語の活用

social

• Platform-optimized (Twitter/X, LinkedIn, Instagram)
• Engaging hooks and calls-to-action
• Emoji usage where natural (but not forced)
• Short, punchy sentences
• For Japanese: SNS向け口語体、略語OK

Meaning-Preserving Transformation Rules

CRITICAL: Never alter the factual content. Only transform:

• Word choice (synonyms, less AI-typical phrasing)
• Sentence structure (vary length, combine or split sentences)
• Transitions (replace formulaic with natural)
• Paragraph flow (add natural topic shifts, personal voice)
• Tone markers (add contractions, rhetorical questions, emphasis)

NEVER:

• Delete substantive content
• Change numbers, dates, or proper nouns
• Introduce claims not in the original
• Remove citations or references
• Change the overall argument or conclusion

Commands

humanize <text-or-file>

1. Detect language
2. Scan for AI patterns
3. Calculate statistical scores
4. Transform text
5. Verify: re-score transformed text (must be lower than original)
6. Show before/after AI score

> humanize "It is important to note that artificial intelligence has significantly
  transformed the landscape of modern technology. Furthermore, the comprehensive
  impact of these advancements cannot be overstated. Moreover, leveraging these
  tools effectively fosters innovation across various domains."

🔍 Language: English | Mode: casual (default)

📊 Before — AI Score: 89%
  Patterns: "important to note", "landscape", "Furthermore", "comprehensive",
  "Moreover", "leveraging", "fosters" (7 ChatGPT-style patterns)

✍ Transformed:
  "AI has changed how we build and use technology — that much is obvious. What's
  less obvious is how deep it goes. These tools aren't just nice to have anymore.
  Teams that actually use them well are shipping faster and finding solutions that
  weren't possible two years ago."

📊 After — AI Score: 14%
  ✅ Score reduced: 89% → 14% (−75 points)
  Meaning preserved: ✓ All factual claims intact

humanize <text> --mode <academic|business|casual|creative|social>

> humanize "The results demonstrate that the proposed methodology yields
  superior outcomes." --mode academic

🔍 Language: English | Mode: academic

📊 Before — AI Score: 72%

✍ Transformed:
  "Results suggest the proposed approach outperforms existing methods, though
  the effect size varies across experimental conditions (see Table 3)."

📊 After — AI Score: 18%
  ✅ Score reduced: 72% → 18% (−54 points)

humanize <text> --lang <code>

> humanize "AIの活用は非常に重要です。包括的なアプローチを踏まえて、
  多角的に検討することが求められます。" --lang ja

🔍 Language: Japanese (forced) | Mode: casual (default)

📊 Before — AI Score: 83%
  パターン: 「非常に重要です」「包括的な」「踏まえて」「多角的に」「求められます」
  文末の均一性: 「〜です」「〜ます」の繰り返し (σ=0.08)

✍ 変換後:
  「AIをどう使うかって、実はすごく大事なんですよね。いろんな角度から
  考えてみると、やれることがまだまだあるなと。」

📊 After — AI Score: 11%
  ✅ スコア低減: 83% → 11% (−72ポイント)
  文末バリエーション: 「〜んですよね」「〜なと」— 自然な口語体

humanize score <text>

╔═══════════════════════════════════════╗
║     AI Detection Score: 73%          ║
╠═══════════════════════════════════════╣
║ Burstiness:    0.23  (low - AI-like) ║
║ Type-Token:    0.78  (high - AI-like)║
║ Perplexity:    low   (AI-like)       ║
║ Transitions:   0.82  (formulaic)     ║
║ Patterns Found: 12 (ChatGPT-style)   ║
╠═══════════════════════════════════════╣
║ Detected patterns:                    ║
║ - "It's important to note" (×2)      ║
║ - "Furthermore" (×3)                 ║
║ - "comprehensive" (×2)               ║
║ - Uniform sentence length (σ=0.12)   ║
╚═══════════════════════════════════════╝

humanize analyze <text>

> humanize analyze "Furthermore, it is essential to consider the multifaceted
  nature of this comprehensive challenge. The landscape continues to evolve
  rapidly, and stakeholders must leverage innovative solutions to foster
  sustainable growth across all sectors."

╔════════════════════════════════════════════════╗
║      Humanize AI Pro — Full Analysis          ║
╠════════════════════════════════════════════════╣
║                                                ║
║ Language: English                              ║
║ Word Count: 32                                 ║
║ Sentence Count: 2                              ║
║                                                ║
║ Statistical Metrics:                           ║
║   Burstiness:     0.18  (low — AI-like)       ║
║   Type-Token:     0.81  (high — AI-like)      ║
║   Perplexity:     low   (AI-like)             ║
║   Transitions:    0.91  (highly formulaic)    ║
║                                                ║
║ Sentence Length Distribution:                  ║
║   Sentence 1: 14 words                        ║
║   Sentence 2: 18 words                        ║
║   Std Dev: 2.83 | Mean: 16.0 | σ/μ = 0.18    ║
║   Human typical range: 0.4-0.8                ║
║                                                ║
║ Model Signature: ChatGPT (87% confidence)     ║
║   Matched tells: "Furthermore", "essential",  ║
║   "multifaceted", "comprehensive",            ║
║   "landscape", "leverage", "foster"           ║
║                                                ║
║ Overall AI Probability: 91%                   ║
║   ████████████████████░░  LIKELY AI-GENERATED  ║
╚════════════════════════════════════════════════╝

humanize batch <directory>

> humanize batch ./blog-drafts/ --mode business

📂 Scanning ./blog-drafts/... 4 files found

[1/4] intro-post.md
  AI Score: 78% → 16% ✅  (saved to ./blog-drafts/intro-post.humanized.md)

[2/4] product-announcement.md
  AI Score: 82% → 19% ✅  (saved to ./blog-drafts/product-announcement.humanized.md)

[3/4] case-study.md
  AI Score: 45% → 12% ✅  (saved to ./blog-drafts/case-study.humanized.md)

[4/4] team-bio.txt
  AI Score: 23% → 18% ✅  (saved to ./blog-drafts/team-bio.humanized.txt)

📊 Batch Summary:
  Files processed: 4/4
  Average score reduction: 65% → 16% (−49 points)
  Mode: business | Language: auto-detected (all English)
  Total processing time: 8.2s

humanize diff <text>

> humanize diff "It is important to note that our comprehensive analysis
  reveals several key findings. Furthermore, these results demonstrate
  the significant impact of our methodology."

╔══════════════════════════════════════════════════════╗
║  Humanize Diff — Changes Highlighted                ║
╠══════════════════════════════════════════════════════╣
║                                                      ║
║  BEFORE (AI Score: 81%):                             ║
║  "[-It is important to note that-] our               ║
║  [-comprehensive-] analysis reveals several key      ║
║  findings. [-Furthermore,-] these results            ║
║  [-demonstrate the significant impact of-] our       ║
║  methodology."                                       ║
║                                                      ║
║  AFTER (AI Score: 13%):                              ║
║  "[+Here's what we found:+] our analysis turned up   ║
║  [+a few things worth+] [+calling out.+] [+The+]     ║
║  results [+show our approach actually made a real+]   ║
║  [+difference+][+.+]"                                ║
║                                                      ║
║  Changes: 5 substitutions, 1 sentence restructured   ║
║  Patterns removed: 4 (important to note,             ║
║    comprehensive, Furthermore, demonstrate the       ║
║    significant impact of)                            ║
║  Meaning preserved: ✓                               ║
╚══════════════════════════════════════════════════════╝

humanize patterns

> humanize patterns "Certainly! Here's a comprehensive overview. It's important
  to note that this landscape is constantly evolving. Furthermore, leveraging
  these tools fosters innovation. In conclusion, the multifaceted nature of
  this challenge requires a nuanced approach."

🔍 AI Pattern Detection Report

Model: ChatGPT (93% confidence)

Detected Patterns (11):
  ┌───────────────────────────────────────────────────┐
  │ Pattern                    │ Type      │ Count    │
  ├───────────────────────────────────────────────────┤
  │ "Certainly!"               │ Opener    │ 1        │
  │ "Here's a comprehensive"   │ Opener    │ 1        │
  │ "important to note"        │ Hedging   │ 1        │
  │ "landscape"                │ Buzzword  │ 1        │
  │ "Furthermore"              │ Transition│ 1        │
  │ "leveraging"               │ Buzzword  │ 1        │
  │ "fosters"                  │ Buzzword  │ 1        │
  │ "In conclusion"            │ Closer    │ 1        │
  │ "multifaceted"             │ Buzzword  │ 1        │
  │ "nuanced"                  │ Buzzword  │ 1        │
  │ Uniform sentence length    │ Structure │ σ=0.09   │
  └───────────────────────────────────────────────────┘

  Category Breakdown:
    Buzzwords: 5 | Transitions: 1 | Openers: 2 | Closers: 1 | Structure: 1
    Hedging: 1

  Recommendation: High AI signal. Use `humanize` to transform.

Japanese-Specific Features

When processing Japanese text:

1. 敬語レベル調整: 文脈に合わせて敬語レベルを自動調整

   • ビジネスメール → 丁寧語ベース
   • ブログ記事 → ですます + 口語混在
   • SNS → 口語中心

2. 文末表現バリエーション: AIの「〜です。〜ます。」の繰り返しを回避

   • 「〜ですよね」「〜なんです」「〜だったりします」
   • 体言止め、倒置法の適度な使用

3. カタカナ語の適正化: 不必要なカタカナ語を日本語に置換

   • 「コンプリヘンシブ」→「包括的」or 文脈で避ける
   • ただし定着したカタカナ語はそのまま維持

4. 漢字/ひらがなバランス: AIは漢字率が高くなりがち

   • 「予め」→「あらかじめ」
   • 「殆ど」→「ほとんど」
   • 自然な開き（ひらがな化）を適用

Why Humanize AI Pro vs Humanize AI?

Comparison with Alternatives

FAQ

Q: How much does it cost in tokens per humanization? A: A typical humanization pass costs 1.5-3x the input text length in tokens. The process involves scanning (input tokens), analysis (200-400 tokens for scoring), transformation (output roughly equal to input length), and verification re-scoring (200-400 tokens). Scoring-only mode (

humanize score

Q: How accurate is the AI detection scoring? A: The scoring system provides a statistical estimate, not a definitive judgment. It combines 5 metrics (burstiness, type-token ratio, perplexity, transition patterns, known AI phrases) for a composite score. Accuracy is highest for unmodified AI outputs (85-90% agreement with commercial detectors) and lower for lightly edited or mixed human-AI text. It should be treated as a guideline, not a guarantee.

Q: Does humanization change the meaning of my text? A: No. Meaning preservation is a core design principle. The skill only transforms word choice, sentence structure, transitions, and tone markers. It explicitly never deletes substantive content, changes numbers/dates/names, introduces new claims, or removes citations. After transformation, the factual content should be identical to the original.

Q: Can I use it for academic papers without risking plagiarism? A: Humanize AI Pro transforms AI-generated text to sound more natural, but it does not address academic integrity concerns. Using AI to generate academic work and then disguising it as human-written may violate your institution's policies regardless of detection scores. The

academic

Q: How does Japanese support compare to English? A: Japanese support is a key differentiator with 300+ language-specific patterns covering keigo levels, sentence-ending variations, katakana overuse, kanji/hiragana balance, and subject omission. English has 500+ patterns but Japanese analysis goes deeper into cultural and stylistic nuances that English-only tools miss entirely. The 5 writing modes each have Japanese-specific transformation rules.

Q: Does it work offline? A: Yes. All pattern detection and statistical analysis happens within your LLM session. No external API calls are made for scoring or transformation. The pattern databases are embedded in the skill instructions. No network access is required at any point.

Q: Can I use Humanize AI Pro with Summarize Pro? A: Yes, this is a common and effective combination. Use Summarize Pro first to condense content, then apply Humanize AI Pro to make the summary sound natural. The pipeline is: source content ->

summarize

humanize

Q: Does it support batch processing of multiple files? A: Yes. Use

humanize batch <directory>

Q: Can it detect which specific AI model generated the text? A: It provides model-specific pattern matching for ChatGPT, Claude, and Gemini. Each model has distinctive tells (e.g., ChatGPT's "Certainly!", Claude's "nuanced", Gemini's "Absolutely!"). The detection reports which model's patterns are most prevalent, but this is probabilistic — heavily edited text or newer model versions may not match known patterns.

Q: How do I customize the transformation strength? A: The 5 writing modes (academic, business, casual, creative, social) control the transformation style rather than strength. For lighter transformations, use

academic

business

casual

social

Error Handling

--lang <code>

--lang

humanize batch