Detects the repetition patterns AI introduces into screenplays — from recycled action lines to entire scenes that accomplish the same dramatic purpose.

## Key Points

- The script feels like it's repeating itself
- Multiple scenes feel like variations of the same conversation
- Action lines describe characters and locations the same way every time
- Written across multiple AI sessions (high amnesia risk)
- After a Scene Function Audit, to catch scenes with duplicate functions
- Two or more scenes where characters discuss the same relationship problem
- Multiple "planning" scenes where the team discusses the same mission
- Repeated "I believe in you" pep talks
- The same argument happening more than once without meaningful escalation
- Characters re-establishing information or emotional dynamics already established
1. Character enters location (action line describing the space)
2. Brief moment of character alone (looking at phone, pouring coffee, staring out window)

## Quick Example

```
Page 5:  Sarah enters. Her dark hair is pulled back in a messy bun.
Page 22: Sarah walks in, dark hair in a messy bun.
Page 47: Sarah appears, hair pulled back in its usual messy bun.
```

```
"sighs" — 14 occurrences
"runs a hand through their hair" — 9 occurrences
"looks away" — 11 occurrences
"clenches jaw" — 8 occurrences
```

Repetition Auditor (Screenplay)

Detects the repetition patterns AI introduces into screenplays — from recycled action lines to entire scenes that accomplish the same dramatic purpose.

When to Use This Skill

• The script feels like it's repeating itself
• Multiple scenes feel like variations of the same conversation
• Action lines describe characters and locations the same way every time
• Written across multiple AI sessions (high amnesia risk)
• After a Scene Function Audit, to catch scenes with duplicate functions

---

Screenplay-Specific Repetition Types

Type 1 — Action Line Recycling

AI describes characters, locations, and actions with identical or near-identical phrasing each time.

What to scan for:

Character introductions that repeat:

Page 5:  Sarah enters. Her dark hair is pulled back in a messy bun.
Page 22: Sarah walks in, dark hair in a messy bun.
Page 47: Sarah appears, hair pulled back in its usual messy bun.

Location descriptions that copy-paste:

Page 3:  INT. DETECTIVE CHEN'S OFFICE - DAY
         Cluttered desk, stacks of case files, a cold cup of coffee.

Page 31: INT. DETECTIVE CHEN'S OFFICE - DAY
         His cluttered desk is covered in case files. A cold cup of
         coffee sits untouched.

Default character business repeated:

"sighs" — 14 occurrences
"runs a hand through their hair" — 9 occurrences
"looks away" — 11 occurrences
"clenches jaw" — 8 occurrences

Build a business inventory: Extract every character's physical actions. A real person in a 110-page script should have 20+ distinct physical behaviors. If they have 5 on rotation, it's AI repetition.

Type 2 — Duplicate Conversations

AI writes the same conversation more than once — characters discussing the same topic, reaching the same conclusion, in scenes that serve the same dramatic purpose.

What to scan for:

• Two or more scenes where characters discuss the same relationship problem
• Multiple "planning" scenes where the team discusses the same mission
• Repeated "I believe in you" pep talks
• The same argument happening more than once without meaningful escalation
• Characters re-establishing information or emotional dynamics already established

The purpose test: Write a one-sentence summary of what each dialogue scene accomplishes. If two scenes have the same summary, one should be cut.

DUPLICATE CONVERSATION MAP:
Scene 12 (p.18): Sarah tells Mark she doesn't trust him → he promises to change
Scene 24 (p.39): Sarah tells Mark she still doesn't trust him → he promises harder
Scene 31 (p.52): Sarah tells Mark trust takes time → he says he understands

DIAGNOSIS: Three scenes with the same function. Consolidate into one
scene where the trust issue ESCALATES, or show the trust issue through
behavior instead of dialogue.

Type 3 — Scene Structure Templates

AI defaults to the same scene skeleton repeatedly.

Common AI scene template:

1. Character enters location (action line describing the space)
2. Brief moment of character alone (looking at phone, pouring coffee, staring out window)
3. Second character enters or calls
4. Conversation
5. One character says something that shifts the dynamic
6. Reaction beat
7. Scene ends on a meaningful look or loaded silence

Diagnostic: Reduce each scene to its structural skeleton. If more than 50% follow the same pattern, the script is templated.

Type 4 — Transition Repetition

AI uses the same transition devices:

Overused transitions:

• SMASH CUT TO: (used for every tonal shift)
• TIME CUT: or "LATER" (only time-passing device)
• Character looks at something → CUT TO the thing they're looking at
• Scene ends on a question → next scene opens with the answer
• "Beat" stage direction appearing every other page

Type 5 — Emotional Beat Repetition

The same emotional arc repeats across different scenes with different characters.

What AI does:

• Character A has a vulnerable moment in scene 8, Character B has an identical vulnerable moment in scene 14
• Three different characters each have a "I never told anyone this, but..." confession scene
• The protagonist has the same "moment of doubt" before every major action

The emotional unique test: List the emotional beats in order. Are there any emotions expressed only ONCE in the script? A good screenplay has several unique emotional moments. An AI screenplay recycles the same 4-5 emotional beats.

Type 6 — Montage Overreliance

AI uses montages to skip the hard work of dramatizing change.

AI montage pattern:

• Training montage (character gets better at something)
• Relationship montage (couple does cute things)
• Investigation montage (detective reads files, makes calls)
• Preparation montage (team gears up)

One montage per script is conventional. Three or more means the AI is using montage as a crutch to avoid writing scenes where change happens through conflict.

---

Scanning Process

Full Script Scan

1. Action line phrase frequency: Extract all action lines, find repeated phrases of 4+ words
2. Scene purpose mapping: One-sentence function for each scene, identify duplicates
3. Character business inventory: All physical actions per character, count frequencies
4. Transition inventory: All transitions used, count frequencies
5. Emotional beat timeline: Map emotional beats, identify repetition
6. Scene skeleton extraction: Structural template for each scene, identify matches

Output Format

# Screenplay Repetition Report
**Title**: [Script title]
**Page count**: [N]

## Action Line Repetition
**Repeated phrases found**: [N]
| Phrase | Count | Pages |
|--------|-------|-------|
| ... | ... | ... |

## Character Business Inventory
| Character | Unique Actions | Most Repeated | Count |
|-----------|---------------|---------------|-------|
| ... | ... | ... | ... |

## Duplicate Scenes
| Scene A | Scene B | Shared Function | Recommendation |
|---------|---------|----------------|----------------|
| ... | ... | ... | Cut/Combine/Escalate |

## Scene Structure Analysis
**Unique structures**: [N] of [total scenes]
**Most common template**: [description]

## Transition Inventory
| Transition | Count | Assessment |
|-----------|-------|-----------|
| ... | ... | ... |

## Revision Priority
[Ordered list of what to fix first]

---

Anti-Patterns

• Flagging intentional callbacks. Screenplays use deliberate repetition (visual motifs, recurring dialogue, bookend scenes). The difference: intentional callbacks EVOLVE. AI repetition is identical.
• Demanding every scene be structurally unique. Some scene types (interrogation, phone call, briefing) have conventional structures. The content should vary even if the form is similar.
• Treating montages as inherently bad. One well-crafted montage per script is a legitimate tool. The problem is when montage replaces dramatization.
• Counting without page proximity. The same phrase 60 pages apart is fine. The same phrase on adjacent pages is a problem. Weight by proximity.