| name | ai-content-detector |
| description | Detect whether page content is AI-generated. Use when user wants to check if a URL, article, or pasted text was written by AI. Analyzes linguistic patterns, structural signals, and vocabulary to produce an AI probability score with detailed breakdown. |
| allowed-tools | Read, Glob, Grep, WebFetch, WebSearch |
AI Content Detector Skill
Analyze text or web page content to determine whether it was likely generated by AI (ChatGPT, Claude, Gemini, etc.). Produces a detailed report with probability score and humanization suggestions.
When to Use
- User provides a URL and asks "is this AI-generated?"
- User pastes text and wants to know if it reads like AI content
- User wants to audit JR Academy pages for AI-sounding content before publishing
- User asks to check content quality / originality
- User says
/ai-content-detector <URL or text>
Input
$ARGUMENTS can be:
- A URL — fetch and analyze the page content
- Pasted text — analyze directly
- A file path — read and analyze the file
Analysis Workflow
Step 1: Obtain Content
- If
$ARGUMENTS is a URL → use WebFetch to retrieve the page content (extract the main article body, ignore nav/footer/ads)
- If
$ARGUMENTS is a file path → use Read to get the file content
- If
$ARGUMENTS is raw text → analyze directly
Step 2: Multi-Dimensional Analysis
Evaluate the content across 8 dimensions. For each dimension, assign a score from 0 (clearly human) to 10 (clearly AI).
Dimension 1: Vocabulary Fingerprints (词汇指纹)
AI models overuse specific words. Check for high density of these:
English AI buzzwords:
- "delve", "tapestry", "landscape", "leverage", "robust", "comprehensive", "streamline"
- "navigate", "foster", "pivotal", "underscores", "multifaceted", "nuanced"
- "cutting-edge", "game-changer", "paradigm shift", "holistic", "synergy"
- "It's important to note", "It's worth mentioning", "It should be noted"
- "In today's [X] landscape", "In the realm of", "When it comes to"
Chinese AI buzzwords (中文 AI 高频词):
- "深入探讨", "全面解析", "值得注意的是", "综上所述", "不容忽视"
- "至关重要", "与日俱增", "应运而生", "不言而喻", "由此可见"
- "在当今...背景下", "随着...的不断发展", "众所周知"
- "总而言之", "一言以蔽之", "毋庸置疑"
- Excessive use of "的" chains (e.g., "...的...的...的...")
Scoring:
- 0-2 buzzwords found → score 1-2
- 3-5 buzzwords found → score 3-5
- 6+ buzzwords found → score 7-10
Dimension 2: Sentence Structure Uniformity (句式均匀度)
AI tends to produce sentences of similar length and structure.
Check for:
- Low variance in sentence length (most sentences 15-25 words)
- Repetitive sentence openers ("This", "The", "It", "Moreover")
- Formulaic transitions between paragraphs
- Parallel structure overuse (every paragraph follows same pattern)
Human writing signals:
- Mix of short punchy sentences and long complex ones
- Sentence length variance > 40%
- Sentence fragments, rhetorical questions, exclamations
- Varied paragraph lengths
Dimension 3: Structural Patterns (结构模式)
AI-generated content has predictable structures:
AI patterns:
- Introduction → 3-5 body sections → conclusion (always)
- Every section has exactly 3 bullet points or sub-points
- Headers follow a perfect hierarchy without exceptions
- Lists are always complete (never "etc." or trailing off)
- Perfect symmetry in section lengths
Human patterns:
- Uneven section lengths
- Tangents and asides
- Some sections more developed than others
- Organic flow vs rigid outline
Dimension 4: Hedging & Qualification (模糊限定)
AI overuses hedging language to appear balanced:
AI hedging patterns:
- "While X, it's also important to consider Y"
- "On one hand... on the other hand..."
- "However, it's crucial to remember that..."
- "That said, there are some considerations..."
- Never takes a strong stance or gives a definitive answer
- Every pro has a con, every advantage a caveat
Human patterns:
- Strong opinions stated directly
- Willingness to be wrong
- Not every claim is qualified
Dimension 5: Personal Voice & Authenticity (个人风格)
AI lacks:
- First-person anecdotes ("When I worked at...", "I once had...")
- Specific named examples (real people, companies, dates)
- Humor, sarcasm, irony
- Colloquialisms, slang, informal language
- Cultural references specific to a region/community
- Typos or minor grammatical imperfections
Human markers:
- "I think", "In my experience", "I've seen teams..."
- Named tools/companies with specific version numbers
- Conversational asides
- Regional expressions
Dimension 6: Content Depth & Specificity (内容深度)
AI tends to:
- Stay surface-level on technical topics
- Give generic examples that could apply to anything
- Avoid specific numbers, dates, version numbers
- Use placeholder-like examples ("Company X", "Project Y")
- Cover topics broadly without going deep
Human markers:
- Specific technical details (version numbers, config snippets)
- Real-world examples with names and dates
- Deep-dive into niche aspects
- Acknowledging edge cases and exceptions
Dimension 7: Conclusion & Summary Patterns (总结模式)
AI conclusion signals:
- "In conclusion," / "To sum up," / "总而言之" / "综上所述"
- Restates every point from the article
- Ends with a generic call-to-action
- "As we move forward..." / "The future of X is bright..."
- "Whether you're a beginner or expert, X is essential..."
Human conclusion:
- May not have an explicit conclusion
- Ends with a thought-provoking question
- Trails off naturally
- May end abruptly with a final strong statement
Dimension 8: Factual Accuracy & Hallucination Signals (事实准确性)
AI hallucination patterns:
- Statistics without sources ("Studies show that 73% of...")
- Vague authority citations ("According to experts...")
- Plausible-sounding but unverifiable claims
- Mixing up similar concepts or dates
Check:
- Are statistics cited with sources?
- Do named references actually exist?
- Are technical details accurate and current?
Step 3: Calculate Overall Score
Overall AI Probability = weighted average of all dimensions
Weights:
- Vocabulary Fingerprints: 20%
- Sentence Uniformity: 15%
- Structural Patterns: 15%
- Hedging & Qualification: 10%
- Personal Voice: 15%
- Content Depth: 10%
- Conclusion Patterns: 5%
- Factual Accuracy: 10%
Step 4: Classify Result
| Score Range | Classification | Label |
|---|
| 0.0 - 2.0 | Very likely human-written | Human |
| 2.1 - 4.0 | Probably human with some AI patterns | Mostly Human |
| 4.1 - 6.0 | Mixed / uncertain | Uncertain |
| 6.1 - 8.0 | Probably AI-generated | Likely AI |
| 8.1 - 10.0 | Very likely AI-generated | AI Generated |
Output Format
# AI Content Detection Report
**Source**: [URL / filename / "pasted text"]
**Content Length**: [X words / Y characters]
**Language**: [English / Chinese / Mixed]
**Date Analyzed**: [today]
---
## Overall Result
| Metric | Value |
|--------|-------|
| **AI Probability Score** | X.X / 10 |
| **Classification** | [Human / Mostly Human / Uncertain / Likely AI / AI Generated] |
| **Confidence** | [High / Medium / Low] |
---
## Dimension Breakdown
| # | Dimension | Score | Key Evidence |
|---|-----------|-------|--------------|
| 1 | Vocabulary Fingerprints | X/10 | [brief evidence] |
| 2 | Sentence Uniformity | X/10 | [brief evidence] |
| 3 | Structural Patterns | X/10 | [brief evidence] |
| 4 | Hedging & Qualification | X/10 | [brief evidence] |
| 5 | Personal Voice | X/10 | [brief evidence] |
| 6 | Content Depth | X/10 | [brief evidence] |
| 7 | Conclusion Patterns | X/10 | [brief evidence] |
| 8 | Factual Accuracy | X/10 | [brief evidence] |
---
## Detected AI Signals
### High-Confidence Signals
- [List specific phrases, patterns, or structures that strongly indicate AI]
### Moderate Signals
- [List moderate indicators]
---
## Detected Human Signals
- [List specific elements that suggest human authorship]
---
## Sample Evidence
> "[Quote 1 from the content that shows AI pattern]"
> — Dimension X signal: [explanation]
> "[Quote 2]"
> — Dimension Y signal: [explanation]
> "[Quote 3]"
> — Dimension Z signal: [explanation]
---
## Humanization Suggestions
If the content scores > 5.0, provide actionable suggestions:
1. **[Suggestion category]**: [Specific change to make]
2. **[Suggestion category]**: [Specific change to make]
3. ...
Example suggestions:
- Replace "delve into" → "dig into" / "look at" / "explore"
- Replace "comprehensive overview" → specific scope description
- Add a personal anecdote in paragraph 2
- Vary sentence length: break up the 20-word average
- Remove hedging in conclusion, state opinion directly
- Add specific data points with sources
- Replace "In today's digital landscape" → specific context
Edge Cases
- Content too short (< 100 words): Report low confidence, note that short texts are harder to analyze
- Technical documentation: Pure API docs / READMEs may score high on structure but are not "AI generated" — adjust for content type
- Mixed content: If parts are clearly human and parts AI, note the boundary and provide per-section analysis
- Translated content: Translation artifacts may mimic AI patterns — flag if translation is suspected
- Heavily edited AI content: Note when content shows signs of AI-assisted but human-edited writing
Quick Examples
Clearly AI (score ~8-9):
"In today's rapidly evolving digital landscape, it's important to delve into the multifaceted world of artificial intelligence. As we navigate the complexities of this transformative technology, it's worth noting that AI has become a pivotal force in reshaping how we approach problem-solving."
Signals: "digital landscape", "delve into", "multifaceted", "navigate the complexities", "pivotal force", "it's worth noting"
Clearly Human (score ~1-2):
"I spent three weeks trying to get GPT-4 to write decent unit tests for our React app. It kept hallucinating import paths. Finally asked my coworker Sarah and she pointed out we were using Vitest, not Jest. Felt dumb but whatever, the tests pass now."
Signals: specific timeframe, named person, self-deprecation, informal tone, specific tools with context, "whatever"
Chinese AI (score ~7-8):
"随着人工智能技术的不断发展，AI 在教育领域的应用日益广泛。值得注意的是，AI 辅助学习不仅提高了学习效率，还为个性化教育提供了全新的可能性。综上所述，拥抱 AI 技术已成为教育工作者不可忽视的趋势。"
Signals: "随着...的不断发展", "值得注意的是", "综上所述", "不可忽视", formulaic structure
