humanizer

Compare original and translation side by side

🇺🇸

Original

English

🇨🇳

Translation

Chinese

Humanizer: AI Pattern Detection & Voice Injection

Humanizer：AI模式检测与真实语气注入

Transform AI-generated text into human writing by detecting patterns and injecting authentic voice.

通过检测AI写作模式并注入真实语气，将AI生成文本转化为人类风格的写作。

Before You Edit: Diagnostic Framework

编辑前：诊断框架

Ask yourself these 3 questions BEFORE applying any patterns:

在应用任何修正模式前，请先问自己这3个问题：

1. Voice Assessment

1. 语气评估

Does this text have a distinct voice? Or is it neutral/corporate?
What personality should come through? (witty, skeptical, conversational, authoritative)
Are there opinions, or just facts? Human writing has stakes and perspective.

这段文本是否有独特的语气？ 还是中立/企业化的风格？
应该展现出怎样的性格？（诙谐、质疑、口语化、权威）
文本中包含观点，还是仅为事实陈述？ 人类写作会带有立场和视角。

2. Pattern Prioritization

2. 模式优先级

Which 3-5 patterns dominate this text? (Don't fix everything at once)
What's the writer's intent? (persuasive → keep some structure; casual → break all patterns)
Should some "AI-isms" stay? (Formal technical docs may keep certain structures)

这段文本中占主导的3-5种AI模式是什么？（不要一次性修正所有问题）
作者的写作意图是什么？（说服性文本→可保留部分结构；非正式文本→打破所有模式）
是否可以保留部分“AI特征”？（正式技术文档可能需要保留特定结构）

3. Rewrite Philosophy

3. 改写原则

Am I removing patterns OR injecting personality? (Must do both)
Does my rewrite sound like a specific human wrote it? (Not just "less AI")
Have I varied sentence rhythm? (Short. Longer flowing sentences. Mix it up.)

The Core Principle: Sterile, voiceless writing is just as obvious as slop. Don't just remove bad patterns—add soul.

我是在去除模式，还是在注入个性？（两者必须同时进行）
改写后的文本听起来像是某个具体的人写的吗？（不只是“不像AI”）
我是否调整了句子节奏？（短句、长句结合，丰富变化）

核心原则：呆板、无个性的写作和粗制滥造的文本一样显眼。不要只去除不良模式——还要注入灵魂。

4. Pattern Detection Procedure (Domain-Specific)

4. 领域专属模式检测流程

Run these checks BEFORE editing:

Statistical Density Check:

Count AI vocabulary words per 100 words: >3 = heavy AI signature
Count em dashes per paragraph: >2 = structural tell
Count "However" paragraph starts: >20% = AI transition overuse

Structural Signature Check:

All paragraphs same length? = AI rhythm uniformity
Every list has exactly 3 items? = rule of three addiction
Conclusions use passive voice? = AI hedging pattern

Context-Specific Preservation:

Academic: Keep formal structure, remove only vocabulary slop
Technical: Preserve precision terminology, remove promotional language
Marketing: Full humanization except brand voice requirements

编辑前请先完成以下检查：

统计密度检查:

每100词中AI词汇数量：>3 = 明显AI特征
每段破折号数量：>2 = 结构特征
以“However”开头的段落占比：>20% = AI过渡语过度使用

结构特征检查:

所有段落长度一致？= AI节奏同质化
每个列表恰好包含3项？= 过度依赖三分法则
结论使用被动语态？= AI的模糊化写作模式

场景特定保留规则:

学术文本：保留正式结构，仅去除冗余词汇
技术文本：保留精准术语，去除宣传性语言
营销文本：除品牌语气要求外，完全人性化处理

Critical Anti-Patterns (NEVER Do This)

关键反模式（绝对不要这么做）

❌ Pattern #1: Mechanical Pattern Removal

❌ 模式1：机械式模式移除

Problem: Just deleting AI phrases without adding human voice produces sterile text.

markdown

❌ BAD EDIT:
"The framework serves as a testament to modern development practices."
→ "The framework is modern."

✅ GOOD EDIT:
"The framework serves as a testament to modern development practices."
→ "This framework gets it right. Clean APIs, sensible defaults, actual documentation."

Why this matters: Removing "testament to" makes it grammatically correct but soulless. The good edit has opinion, rhythm, and personality.

问题：仅删除AI短语但不注入人类语气，会产生呆板的文本。

markdown

❌ 错误编辑:
"The framework serves as a testament to modern development practices."
→ "The framework is modern."

✅ 正确编辑:
"The framework serves as a testament to modern development practices."
→ "This framework gets it right. Clean APIs, sensible defaults, actual documentation."

重要性：删除“testament to”虽然语法正确，但毫无灵魂。正确的编辑包含观点、节奏和个性。

❌ Pattern #2: Over-Correction

❌ 模式2：过度修正

Problem: Making every sentence "unpredictable" creates chaos, not humanity.

markdown

❌ BAD EDIT (too chaotic):
"Results. Interesting ones! The experiment? It generated code—lots of it.
3 million lines worth. Developers (some of them) were impressed!!!!"

✅ GOOD EDIT (controlled variety):
"I genuinely don't know how to feel about this. 3 million lines of code,
generated overnight. Half the dev community is losing their minds,
half are explaining why it doesn't count."

Why this matters: Human writing has rhythm variation, not random punctuation chaos.

问题：让每个句子都“不可预测”会造成混乱，而非人性化。

markdown

❌ 错误编辑（过于混乱）:
"Results. Interesting ones! The experiment? It generated code—lots of it.
3 million lines worth. Developers (some of them) were impressed!!!!"

✅ 正确编辑（可控的多样性）:
"I genuinely don't know how to feel about this. 3 million lines of code,
generated overnight. Half the dev community is losing their minds,
half are explaining why it doesn't count."

重要性：人类写作有节奏变化，而非随机的标点混乱。

❌ Pattern #3: Removing ALL Structure

❌ 模式3：移除所有结构

Problem: Not all AI patterns are bad—some formal writing needs structure.

markdown

Context: Academic paper abstract

❌ BAD EDIT:
"Our study looked at machine learning. We found some stuff.
It's interesting. Check out our results."

✅ GOOD EDIT:
"This study examines machine learning approaches to code generation.
We evaluated three architectures and found that transformer-based
models outperformed RNNs by 23% on our benchmark."

Why this matters: Formal contexts need clarity over personality. Know your audience.

问题：并非所有AI模式都是坏的——部分正式写作需要结构支撑。

markdown

场景：学术论文摘要

❌ 错误编辑:
"Our study looked at machine learning. We found some stuff.
It's interesting. Check out our results."

✅ 正确编辑:
"This study examines machine learning approaches to code generation.
We evaluated three architectures and found that transformer-based
models outperformed RNNs by 23% on our benchmark."

重要性：正式场景需要清晰性而非个性。要了解你的受众。

❌ Pattern #4: Batch-Replacing AI Words Without Context

❌ 模式4：无上下文批量替换AI词汇

Problem: Blindly replacing "delve" or "landscape" breaks legitimate usage.

markdown

Context: Computer vision paper

❌ BAD EDIT:
"Our model examines the feature landscape" → "Our model examines the feature terrain"

✅ GOOD EDIT:
"Our model examines the feature landscape" → "Our model analyzes feature space"
OR keep "landscape" if it's established terminology in CV papers

Why this matters: Not every AI word is wrong—check if it's domain-appropriate first. "Landscape" in data science ≠ "business landscape" slop.

问题：盲目替换“delve”或“landscape”会破坏合理用法。

markdown

场景：计算机视觉论文

❌ 错误编辑:
"Our model examines the feature landscape" → "Our model examines the feature terrain"

✅ 正确编辑:
"Our model examines the feature landscape" → "Our model analyzes feature space"
或如果“landscape”是CV领域通用术语则保留

重要性：并非所有AI词汇都是错误的——先确认是否符合领域规范。数据科学中的“landscape”≠ 商业文本中的冗余表述。

Most Common AI Patterns (Quick Reference)

最常见的AI模式（速查）

Content-Level Patterns

内容层面模式

Undue Emphasis on Significance

Words: stands as, serves as, testament to, pivotal, crucial, underscores, broader trends
Fix: Remove inflated symbolism, state facts directly

Promotional Language

Words: boasts, nestled, vibrant, rich heritage, breathtaking, stunning
Fix: Replace adjectives with specific details

Vague Attributions

Words: Industry reports, Observers note, Experts argue, Some critics
Fix: Name specific sources or remove the claim

过度强调重要性

词汇：stands as、serves as、testament to、pivotal、crucial、underscores、broader trends
修正：去除夸张表述，直接陈述事实

宣传性语言

词汇：boasts、nestled、vibrant、rich heritage、breathtaking、stunning
修正：用具体细节替换形容词

模糊归因

词汇：Industry reports、Observers note、Experts argue、Some critics
修正：指明具体来源或删除该表述

Language-Level Patterns

语言层面模式

AI Vocabulary Words (post-2023 frequency spike)

Words: delve, crucial, enhance, foster, garner, intricate, landscape (abstract), pivotal, showcase, tapestry (abstract), underscore
Fix: Use plain synonyms or restructure

Copula Avoidance (avoiding "is/are")

Pattern: "serves as", "stands as", "represents", "boasts", "features"
Fix: Use simple "is/are/has"

Negative Parallelisms

Pattern: "Not only... but...", "It's not just about X, it's Y"
Fix: State directly without forced contrast

AI高频词汇（2023年后使用量激增）

词汇：delve、crucial、enhance、foster、garner、intricate、landscape（抽象用法）、pivotal、showcase、tapestry（抽象用法）、underscore
修正：使用直白同义词或重构句子

避免系动词（避免使用“is/are”）

模式："serves as"、"stands as"、"represents"、"boasts"、"features"
修正：使用简单的"is/are/has"

否定平行结构

模式："Not only... but..."、"It's not just about X, it's Y"
修正：直接陈述，无需刻意对比

Style-Level Patterns

风格层面模式

Em Dash Overuse

Pattern: Multiple em dashes in one paragraph (—)
Fix: Replace with commas, periods, or parentheses

Rule of Three Overuse

Pattern: "innovation, inspiration, and industry insights"
Fix: Break groups of three, vary list sizes

Title Case Headings

Pattern: "Strategic Negotiations And Global Partnerships"
Fix: Sentence case: "Strategic negotiations and global partnerships"

过度使用破折号

模式：单段中出现多个破折号（—）
修正：替换为逗号、句号或括号

过度依赖三分法则

模式："innovation, inspiration, and industry insights"
修正：打破三组结构，丰富列表长度

标题大小写标题

模式："Strategic Negotiations And Global Partnerships"
修正：句子大小写："Strategic negotiations and global partnerships"

Humanization Strategy: When to Preserve vs Remove

人性化策略：保留与移除的判断

The Decision Framework: Not all contexts need full humanization.

Context	Humanization Level	Remove Patterns	Inject Voice	Example Fix
Academic/Research	Low (10-20%)	Delete slop only (delve, testament to)	Minimal	Keep structure, remove AI vocabulary
Technical Docs	Medium (30-50%)	Remove promotional language, keep clarity	Light opinions	"This works well" → "This approach handles edge case X"
Blog/Marketing	High (70-90%)	Remove most AI tells	Strong voice	Full personality, distinct author presence
Social/Casual	Maximum (100%)	Delete all AI patterns	Maximum authenticity	Pure conversational, break all rules
Formal Business	Medium (40-60%)	Remove obvious slop, keep professionalism	Controlled confidence	"We believe this represents..." → "This delivers X"

Critical Non-Obvious AI Tells (beyond the common list):

Paragraph-starting "However": AI overuses this transition (appears 3x more in GPT text)
Passive voice in conclusions: "It can be concluded that..." (AI hedges at the end)
Symmetric sentence structure: Every paragraph follows same length/rhythm pattern
"Importantly" mid-sentence: AI uses this more than humans (statistical quirk)
Abstract "landscape" metaphors: "the technology landscape", "the business landscape"

决策框架：并非所有场景都需要完全人性化。

场景	人性化程度	需移除模式	需注入语气	修正示例
学术/研究	低（10-20%）	仅删除冗余内容（delve、testament to）	极少	保留结构，移除AI词汇
技术文档	中（30-50%）	移除宣传性语言，保留清晰性	轻度观点	"This works well" → "This approach handles edge case X"
博客/营销	高（70-90%）	移除大部分AI特征	强烈语气	完整个性，突出作者风格
社交/非正式	最高（100%）	删除所有AI模式	最大程度真实	纯口语化，打破所有规则
正式商务	中（40-60%）	移除明显冗余，保留专业性	可控的自信语气	"We believe this represents..." → "This delivers X"

易被忽略的AI特征（超出常见列表）：

以"However"开头的段落：AI过度使用该过渡词（GPT文本中出现频率是人类的3倍）
结论使用被动语态："It can be concluded that..."（AI在结尾的模糊化表述）
对称句子结构：每段遵循相同长度/节奏模式
句中使用"Importantly"：AI使用频率高于人类（统计特征）
抽象"landscape"隐喻："the technology landscape"、"the business landscape"

When to Load Full Pattern References

何时加载完整模式参考

For comprehensive pattern catalogs, use mandatory loading:

MANDATORY - READ ENTIRE FILE:
references/content-patterns.md
when:

Text contains 5+ promotional adjectives (stunning, breathtaking, vibrant, rich)
Significance inflation detected ("serves as testament", "stands as pivotal")
Need complete "symbolism removal" examples
Do NOT load for casual blog posts or social media text

MANDATORY - READ ENTIRE FILE:
references/language-patterns.md
when:

Text uses 8+ AI vocabulary words (delve, showcase, intricate, foster, garner)
Heavy copula avoidance patterns ("serves as" instead of "is")
Need elegant variation catalog for substitutions
Do NOT load for technical documentation where precision matters

MANDATORY - READ ENTIRE FILE:
references/style-patterns.md
when:

Text has 6+ em dashes in single paragraph
Rule of three appears 4+ times
Title case headings throughout document
Do NOT load for academic papers (formatting may be required)

Never load references for simple opinion injection or rhythm fixes—handle with decision framework above.

如需全面的模式目录，请按以下规则强制加载：

强制要求 - 完整阅读文件：
references/content-patterns.md
当：

文本包含5个以上宣传性形容词（stunning、breathtaking、vibrant、rich）
检测到夸张表述（"serves as testament"、"stands as pivotal"）
需要完整的“象征主义移除”示例
请勿加载于非正式博客或社交媒体文本

强制要求 - 完整阅读文件：
references/language-patterns.md
当：

文本使用8个以上AI词汇（delve、showcase、intricate、foster、garner）
存在大量避免系动词的模式（用"serves as"代替"is"）
需要优雅的替换词目录
请勿加载于需要精准表述的技术文档

强制要求 - 完整阅读文件：
references/style-patterns.md
当：

单段包含6个以上破折号
三分法则出现4次以上
文档全程使用标题大小写标题
请勿加载于学术论文（可能有格式要求）

仅需注入观点或调整节奏时，请勿加载参考文件——使用上述决策框架处理即可。

Process

操作流程

Read input text - Identify 3-5 dominant patterns
Apply diagnostic framework - Answer the 3 questions above
Make strategic edits - Fix patterns + inject voice simultaneously
Verify rhythm - Read aloud test (does it sound natural?)
Present result - Show rewritten text with brief summary if helpful

阅读输入文本 - 识别3-5种主导模式
应用诊断框架 - 回答上述3个问题
进行策略性编辑 - 同时修复模式并注入语气
验证节奏 - 朗读测试（听起来自然吗？）
呈现结果 - 展示改写后的文本，如有需要可附简短说明

Quick Example

快速示例

Before (AI-sounding):

The new software update serves as a testament to the company's commitment to innovation. Moreover, it provides a seamless, intuitive, and powerful user experience—ensuring that users can accomplish their goals efficiently. It's not just an update, it's a revolution in how we think about productivity.

After (Humanized):

The software update adds batch processing, keyboard shortcuts, and offline mode. Early beta feedback has been positive—most testers report finishing tasks faster.

What changed:

Removed inflated symbolism ("serves as a testament")
Removed AI vocabulary ("Moreover", "seamless, intuitive, powerful")
Removed negative parallelism ("It's not just...it's...")
Removed vague claims ("commitment to innovation")
Added specific features (batch processing, shortcuts, offline)
Added concrete evidence (beta feedback, faster completion)
Kept neutral tone appropriate for feature announcement

改写前（AI风格）：

The new software update serves as a testament to the company's commitment to innovation. Moreover, it provides a seamless, intuitive, and powerful user experience—ensuring that users can accomplish their goals efficiently. It's not just an update, it's a revolution in how we think about productivity.

改写后（人性化风格）：

这次软件更新新增了批量处理、快捷键和离线模式。早期beta测试反馈不错——大多数测试者表示任务完成速度更快了。

改动说明：

移除了夸张表述（"serves as a testament"）
移除了AI词汇（"Moreover"、"seamless, intuitive, powerful"）
移除了否定平行结构（"It's not just...it's..."）
移除了模糊表述（"commitment to innovation"）
添加了具体功能（批量处理、快捷键、离线模式）
添加了具体证据（beta反馈、更快完成任务）
保留了适合功能公告的中立语气

Reference Materials

参考资料

Based on Wikipedia:Signs of AI writing, maintained by WikiProject AI Cleanup.

Key insight: "LLMs use statistical algorithms to guess what should come next. The result tends toward the most statistically likely result that applies to the widest variety of cases."

Translation: AI writing is optimized for average acceptability, not authentic voice.

基于维基百科条目Wikipedia:Signs of AI writing，由WikiProject AI Cleanup维护。

核心观点：“大语言模型（LLM）使用统计算法预测下一个内容，结果倾向于适用于最广泛场景的统计最优解。”

通俗解释：AI写作是为了普适性接受度优化，而非真实语气。