Detects every form of repetition that AI introduces into novels — from obvious phrase duplication to subtle structural and thematic loops that make a manuscript feel like it's running in place.

## Key Points

- User says "this feels repetitive" or "I keep reading the same thing"
- Manuscript was written across multiple AI sessions (high amnesia risk)
- Novel is 60,000+ words and needs cross-chapter repetition analysis
- After an AI Tell Detector pass, to catch the subtler repetition layer beneath phrase-level tells
- Identical phrases of 4+ words appearing more than once
- Near-identical phrases (same structure, 1-2 words swapped): "her breath caught in her throat" / "his breath caught in his chest"
- Pet phrases the AI returns to across sessions: "the weight of it all," "something unreadable in their eyes," "the silence that followed"
- Same phrase appearing 2x in a 80K-word novel: acceptable (some repetition is natural)
- Same phrase appearing 3-4x: flag as pattern
- Same phrase appearing 5+: critical — reads as broken record
- Character introductions that repeat the same physical details each appearance ("her dark hair cascading over her shoulders" in chapters 1, 4, 7, 11, 15)
- Settings described with the same details in the same order every time visited

Repetition Auditor

Detects every form of repetition that AI introduces into novels — from obvious phrase duplication to subtle structural and thematic loops that make a manuscript feel like it's running in place.

When to Use This Skill

• User says "this feels repetitive" or "I keep reading the same thing"
• Manuscript was written across multiple AI sessions (high amnesia risk)
• Novel is 60,000+ words and needs cross-chapter repetition analysis
• After an AI Tell Detector pass, to catch the subtler repetition layer beneath phrase-level tells

The Repetition Taxonomy

Level 1 — Exact and Near-Exact Phrase Repetition

The most obvious form. AI reuses the same phrases because it has no memory of having used them before.

What to scan for:

• Identical phrases of 4+ words appearing more than once
• Near-identical phrases (same structure, 1-2 words swapped): "her breath caught in her throat" / "his breath caught in his chest"
• Pet phrases the AI returns to across sessions: "the weight of it all," "something unreadable in their eyes," "the silence that followed"

Scoring:

• Same phrase appearing 2x in a 80K-word novel: acceptable (some repetition is natural)
• Same phrase appearing 3-4x: flag as pattern
• Same phrase appearing 5+: critical — reads as broken record

Output format:

PHRASE REPETITION MAP
"a flicker of something" — 7 occurrences
  Ch. 2 (p.14), Ch. 5 (p.38), Ch. 8 (p.61), Ch. 12 (p.89),
  Ch. 15 (p.112), Ch. 19 (p.141), Ch. 22 (p.167)
  Severity: CRITICAL — appears every ~11,000 words
  Fix: Keep the strongest instance. Replace others with scene-specific detail.

Level 2 — Description Recycling

AI describes the same character, setting, or object nearly identically each time it appears. Real authors vary descriptions based on context, mood, and whose POV is active.

What to scan for:

• Character introductions that repeat the same physical details each appearance ("her dark hair cascading over her shoulders" in chapters 1, 4, 7, 11, 15)
• Settings described with the same details in the same order every time visited
• Weather and atmosphere descriptions that recycle: "the sun dipped below the horizon" appearing at every scene transition
• Food, clothing, and object descriptions that use identical language

Diagnostic questions:

• Does the character's description evolve or stay frozen at their first appearance?
• Does the setting look different at different times of day, seasons, or emotional contexts?
• Are descriptions filtered through the current POV character's perspective, or are they narrator-generic?

Level 3 — Reaction Repetition

AI gives characters the same emotional reaction to different situations. This is one of the most damaging forms because it flattens character.

What to scan for:

• Same physical reaction to different emotions: character "swallows hard" when nervous, angry, sad, and surprised
• Same internal thought pattern: character "wonders if things will ever be the same" after every setback
• Same dialogue response: character responds to conflict by saying "I just need some time" regardless of the situation
• Emotional reactions that don't scale: character reacts to a minor inconvenience with the same intensity as a major trauma

Build a reaction inventory per character:

CHARACTER: Sarah
  Stress reaction: "clenched her jaw" — 11 occurrences
  Sadness reaction: "blinked back tears" — 8 occurrences
  Surprise reaction: "eyes widened" — 14 occurrences
  Anger reaction: "hands curled into fists" — 9 occurrences

  DIAGNOSIS: Sarah has 4 physical reactions total. A real person has dozens.
  She clenches her jaw whether her coffee is cold or her mother dies.

Level 4 — Structural Repetition

AI defaults to the same scene structure, chapter shape, and narrative rhythm.

What to scan for:

• Chapters that follow the same template: open with setting → dialogue exchange → internal reflection → cliffhanger
• Scenes that always resolve in the same pattern: conflict → momentary tension → someone says something wise → resolution
• Transitions that always use the same device: time skips, weather changes, or "the next morning"
• Paragraph structure that repeats: action-beat/dialogue/thought, action-beat/dialogue/thought
• Every chapter being within 500 words of the same length

Structural rhythm map:

Chapter lengths: 3200, 3100, 3400, 3000, 3300, 3200, 3100...
  DIAGNOSIS: Pathologically uniform. Real novels vary chapter length
  by 50-200%. Range here is only 13%.

Scene openings: 18 of 24 chapters open with the character waking up,
  looking out a window, or arriving somewhere.
  DIAGNOSIS: 75% same opening device. Vary radically.

Level 5 — Thematic Hammering

AI states its themes explicitly and repeatedly, not trusting the reader to absorb them through story.

What to scan for:

• The same moral or lesson stated in narration multiple times: "she realized that home wasn't a place — it was the people you loved" appearing as narration in chapters 3, 9, 14, and 22
• Characters voicing the theme in dialogue: multiple characters independently arrive at the same philosophical conclusion
• The narrator summarizing what a scene "means" after the scene already showed it
• Epigraphs, chapter titles, or section breaks that restate the theme explicitly

The hammering test: Remove every explicit statement of the theme. Is the theme still clear from the story events alone? If yes, the explicit statements are redundant. If no, the story needs work, not more theme statements.

Level 6 — Cross-Chapter Amnesia

AI literally forgets what happened and re-introduces information, backstory, or exposition that was already covered.

What to scan for:

• Backstory dumps that repeat: character's tragic history explained in chapter 3 AND chapter 11
• Worldbuilding exposition restated: the rules of the magic system explained twice
• Relationship dynamics re-established: "they hadn't spoken since the argument" stated in three different chapters
• Character introductions repeated: secondary character gets a full re-introduction as if appearing for the first time
• Previously resolved plot points revisited as if unresolved

This is the most detectable AI pattern for attentive readers. It's the equivalent of someone telling you the same story twice at a party.

---

Scanning Process

Full Manuscript Scan

1. Phrase frequency analysis: Extract all phrases of 4-8 words, count occurrences, flag any appearing 3+ times
2. Description comparison: For each recurring character/setting, extract all descriptions and compare side-by-side
3. Reaction inventory: Build per-character reaction tables
4. Structural analysis: Map chapter lengths, opening devices, scene structures, transition types
5. Theme extraction: Identify explicit theme statements and map their locations
6. Amnesia check: Flag any expository information that appears more than once

Output Format

# Repetition Audit Report
**Title**: [Novel title]
**Word count**: [N]
**Date**: [Today]

## Executive Summary
**Overall repetition severity**: [Low / Moderate / High / Critical]
**Estimated revision scope**: [N] passages need rewriting
**Most damaged area**: [which repetition type is worst]

## Level 1: Phrase Repetition
**Unique repeated phrases found**: [N]
**Total excess occurrences**: [N]

| Phrase | Count | Chapters | Severity |
|--------|-------|----------|----------|
| ... | ... | ... | ... |

## Level 2: Description Recycling
**Characters with recycled descriptions**: [N]
**Settings with recycled descriptions**: [N]

[Per-character/setting comparison tables]

## Level 3: Reaction Repetition
[Per-character reaction inventories with counts]

## Level 4: Structural Repetition
**Chapter length variance**: [percentage]
**Scene opening diversity**: [N unique openings / N total chapters]
**Dominant scene structure**: [description]

## Level 5: Thematic Hammering
**Explicit theme statements found**: [N]
**Unique themes**: [N]
**Most hammered theme**: "[theme]" — stated [N] times

## Level 6: Cross-Chapter Amnesia
**Repeated exposition blocks**: [N]
**Repeated backstory dumps**: [N]
**Re-introduced characters**: [N]

## Revision Priority Queue
[Ordered list of what to fix first, with specific guidance for each]

---

Anti-Patterns

• Flagging intentional repetition. Literary repetition (motifs, refrains, callback imagery) is a deliberate craft technique. The difference: intentional repetition evolves and deepens with each recurrence. AI repetition is flat — identical each time.
• Counting without context. "Said" appearing 200 times is not repetition — it's invisible dialogue attribution. Focus on content words and phrases, not function words.
• Treating all repetition as equal. A repeated metaphor in adjacent paragraphs is worse than the same metaphor 40,000 words apart. Weight by proximity.
• Offering generic fixes. "Vary your descriptions" is useless advice. Provide specific alternatives grounded in the scene: "In chapter 11, describe Sarah through Marcus's eyes — he notices different things than the narrator does in chapter 1."
• Missing the forest for the trees. Sometimes the problem isn't that phrases repeat — it's that scenes repeat. Two scenes that accomplish the same narrative purpose should be merged, not just reworded.