F010 - AI Analysis Methods (AutoCrit Feature Parity)
Objective
Document comprehensive AI-only analysis methods that match and exceed AutoCrit’s manuscript analysis capabilities using OpenAI/Claude without external databases or human editors.
AutoCrit Analysis Categories - AI Implementation
1. Pacing & Momentum Analysis
How AutoCrit Does It
- Analyzes sentence variation, paragraph variation, chapter variation
- Identifies slow-paced paragraphs compared to genre standards
- Provides pacing comparison with bestselling novels
Our AI-Only Implementation
interface PacingAnalysis {
sentenceVariation: {
score: number; // 0-100
averageLength: number;
variationCoefficient: number;
shortSentenceRatio: number; // <10 words
longSentenceRatio: number; // >25 words
examples: {
effective: string[];
needsImprovement: string[];
};
};
paragraphFlow: {
score: number;
averageParagraphLength: number;
dialogueToNarrativeRatio: number;
transitionEffectiveness: number;
slowSections: Array<{
location: string;
reason: string;
suggestion: string;
}>;
};
chapterPacing: {
score: number;
chapterLengthConsistency: number;
endingHooks: number;
openingStrength: number;
pacingVariation: number;
};
overallMomentum: {
score: number;
tensionCurve: Array<{ chapter: number; tensionLevel: number }>;
actionSceneRatio: number;
contemplativeSceneRatio: number;
};
}
// AI Prompt Strategy for Pacing Analysis
const PACING_ANALYSIS_PROMPT = `
You are a professional manuscript editor analyzing pacing and momentum.
ANALYSIS REQUIREMENTS:
1. SENTENCE VARIATION: Count tokens per sentence and evaluate variety
- Short sentences (1-10 words): Good for impact, tension
- Medium sentences (11-25 words): Standard narrative flow
- Long sentences (25+ words): Description, complex ideas
- Flag monotonous patterns
2. PARAGRAPH STRUCTURE: Evaluate paragraph flow and rhythm
- Dialogue-heavy paragraphs: Fast pace
- Description-heavy paragraphs: Slower pace
- Action paragraphs: Dynamic pace
- Identify paragraphs that feel sluggish
3. CHAPTER MOMENTUM: Assess chapter-level pacing
- Opening hooks and closing hooks
- Balance of action vs. contemplation
- Scene transitions and their effectiveness
4. SLOW SECTION IDENTIFICATION: Find areas that drag
- Long descriptive passages without action
- Repetitive internal monologue
- Lack of conflict or tension
- Poor dialogue-to-narrative balance
MANUSCRIPT CONTENT: [CONTENT_HERE]
Provide detailed analysis with specific examples and numerical scores.
`;
async function analyzePacing(content: string, genre: string): Promise<PacingAnalysis> {
const prompt = buildPacingPrompt(content, genre);
const aiResponse = await aiService.analyze(prompt);
return {
sentenceVariation: extractSentenceMetrics(aiResponse),
paragraphFlow: extractParagraphMetrics(aiResponse),
chapterPacing: extractChapterMetrics(aiResponse),
overallMomentum: calculateMomentumScore(aiResponse)
};
}2. Dialogue Analysis
How AutoCrit Does It
- Checks for natural-sounding dialogue
- Identifies repetitive dialogue tags
- Flags unnecessary adverbs in dialogue
- Evaluates character voice distinction
Our AI-Only Implementation
interface DialogueAnalysis {
naturalness: {
score: number;
conversationalFlow: number;
authenticity: number;
examples: {
natural: string[];
stilted: string[];
};
};
characterVoice: {
score: number;
voiceDistinction: number;
consistencyPerCharacter: Record<string, number>;
speechPatterns: Record<string, string[]>;
};
technicalQuality: {
score: number;
tagVariety: number;
adverbOveruse: number;
punctuationCorrectness: number;
repetitiveTagCount: number;
};
subtext: {
score: number;
subtextPresence: number;
emotionalDepth: number;
conflictInDialogue: number;
};
}
const DIALOGUE_ANALYSIS_PROMPT = `
You are analyzing dialogue quality in this manuscript. Evaluate:
1. NATURALNESS: Does dialogue sound like real conversation?
- Contractions usage (natural speech patterns)
- Sentence fragments and interruptions
- Realistic vocabulary for character age/background
- Avoid overly formal or stiff language
2. CHARACTER VOICE DISTINCTION: Can you tell characters apart by speech alone?
- Unique vocabulary per character
- Different sentence structures
- Consistent speech patterns
- Age/background-appropriate language
3. TECHNICAL DIALOGUE CRAFT:
- Count dialogue tags: "said" vs. other tags
- Identify adverbs in dialogue tags (usually unnecessary)
- Check punctuation correctness
- Flag repetitive tag patterns
4. SUBTEXT AND DEPTH:
- Characters saying one thing, meaning another
- Emotional undercurrents in conversation
- Conflict and tension in dialogue
- Characters revealing personality through speech
DIALOGUE EXAMPLES FROM MANUSCRIPT: [DIALOGUE_HERE]
Provide specific feedback with examples of effective and problematic dialogue.
`;3. Character Development Analysis
How AutoCrit Does It
- Tracks character arcs and development
- Identifies lack of clear motivation
- Checks for character consistency
- Analyzes protagonist development
Our AI-Only Implementation
interface CharacterAnalysis {
arcDevelopment: {
score: number;
protagonistArc: {
startingPoint: string;
growthMoments: string[];
endingPoint: string;
changeMetrics: number;
};
supportingCharacterArcs: Array<{
character: string;
arcScore: number;
development: string;
}>;
};
motivation: {
score: number;
goalsClarity: number;
motivationConsistency: number;
internalConflict: number;
externalConflict: number;
};
consistency: {
score: number;
behaviorConsistency: number;
speechPatternConsistency: number;
physicalDescriptionConsistency: number;
contradictions: string[];
};
relationshipDynamics: {
score: number;
relationshipDevelopment: number;
conflictDynamics: number;
supportingRelationships: number;
};
}
const CHARACTER_ANALYSIS_PROMPT = `
Analyze character development throughout this manuscript:
1. CHARACTER ARC TRACKING:
- Identify the protagonist and their starting emotional/psychological state
- Track key moments of growth, change, or realization
- Evaluate the ending state vs. beginning state
- Assess if the character truly changes and grows
2. MOTIVATION ANALYSIS:
- What does each major character want? (external goals)
- What do they need? (internal growth)
- Are motivations clear and compelling?
- Do characters make decisions consistent with their goals?
3. CONSISTENCY CHECKING:
- Physical descriptions: height, eye color, age consistency
- Personality traits: behavior patterns throughout story
- Speech patterns: vocabulary and manner of speaking
- Skills and abilities: what characters can/cannot do
4. RELATIONSHIP DYNAMICS:
- How do character relationships evolve?
- Are conflicts between characters realistic and compelling?
- Do characters influence each other's growth?
MANUSCRIPT CONTENT: [CONTENT_HERE]
Focus on major characters and provide specific examples of development or lack thereof.
`;4. Plot Structure Analysis
How AutoCrit Does It
- Identifies plot holes and contradictory events
- Analyzes plot structure and development
- Checks for unresolved plot threads
- Evaluates conflict escalation
Our AI-Only Implementation
interface PlotAnalysis {
structure: {
score: number;
threeActStructure: {
act1Percentage: number;
act2Percentage: number;
act3Percentage: number;
structureBalance: number;
};
plotPoints: {
incitingIncident: string;
midpoint: string;
climax: string;
resolution: string;
};
};
consistency: {
score: number;
plotHoles: Array<{
issue: string;
location: string;
severity: 'minor' | 'major' | 'critical';
}>;
contradictions: string[];
timelineConsistency: number;
};
conflictDevelopment: {
score: number;
conflictEscalation: number;
stakesProgression: number;
resolutionSatisfaction: number;
};
threadResolution: {
score: number;
unresolvedThreads: string[];
setupPayoffPairs: Array<{
setup: string;
payoff: string;
effectiveness: number;
}>;
};
}
const PLOT_ANALYSIS_PROMPT = `
Analyze plot structure and consistency:
1. STORY STRUCTURE:
- Identify Act 1 (setup): approximately 25% of story
- Identify Act 2 (development): approximately 50% of story
- Identify Act 3 (resolution): approximately 25% of story
- Locate: inciting incident, midpoint twist, climax, resolution
2. PLOT HOLE DETECTION:
- Timeline inconsistencies (character ages, dates, seasons)
- Character ability inconsistencies (suddenly knowing skills)
- Logic gaps (how did character get from A to B?)
- Missing motivation explanations
- Contradictory information between scenes
3. CONFLICT ESCALATION:
- Does tension increase throughout the story?
- Are stakes raised appropriately?
- Do obstacles become more challenging?
- Is the climax the highest point of tension?
4. SETUP AND PAYOFF:
- Information introduced early that becomes important later
- Chekhov's gun principle: if introduced, must be used
- Foreshadowing effectiveness
- Unresolved plot threads
MANUSCRIPT CONTENT: [CONTENT_HERE]
Identify specific issues with exact locations and severity levels.
`;5. Point of View (POV) Analysis
How AutoCrit Does It
- Checks POV consistency throughout manuscript
- Identifies head-hopping issues
- Evaluates POV effectiveness
Our AI-Only Implementation
interface POVAnalysis {
consistency: {
score: number;
povType: 'first' | 'second' | 'third-limited' | 'third-omniscient' | 'mixed';
povSwitches: Array<{
location: string;
fromCharacter: string;
toCharacter: string;
appropriate: boolean;
}>;
headHoppingInstances: string[];
};
effectiveness: {
score: number;
povChoiceAppropriate: number;
characterVoiceStrength: number;
intimacyLevel: number;
};
}
const POV_ANALYSIS_PROMPT = `
Analyze point of view consistency and effectiveness:
1. POV IDENTIFICATION:
- Identify primary POV type (first person, third limited, etc.)
- Track which character's perspective we're following
- Note any POV switches between scenes/chapters
2. CONSISTENCY CHECKING:
- Can we only know what the POV character knows?
- Do we only see what the POV character can see?
- Are there inappropriate "mind reading" moments?
- Is internal voice consistent with POV character?
3. HEAD-HOPPING DETECTION:
- Switching POV within a scene without clear breaks
- Knowing other characters' thoughts inappropriately
- Seeing things the POV character couldn't see
MANUSCRIPT CONTENT: [CONTENT_HERE]
Flag specific instances of POV violations with exact quotes.
`;6. World Building Analysis
How AutoCrit Does It
- Evaluates world-building consistency
- Analyzes setting development
- Checks for authentic details
Our AI-Only Implementation
interface WorldBuildingAnalysis {
consistency: {
score: number;
settingDetails: number;
rulesConsistency: number;
geographyLogic: number;
};
immersion: {
score: number;
sensoryDetails: number;
culturalAuthenticity: number;
believability: number;
};
development: {
score: number;
worldComplexity: number;
originalityScore: number;
integrationWithPlot: number;
};
}
// Additional analysis methods for Strong Writing, Word Choice, and Foreshadowing
// follow similar patterns with specific AI prompts and scoring algorithms7. AI Response Processing
class AIResponseProcessor {
parseAnalysisResponse(response: string, category: string): CategoryResult {
try {
const parsed = JSON.parse(response);
return {
score: this.validateScore(parsed.score),
feedback: parsed.feedback || '',
strengths: parsed.strengths || [],
weaknesses: parsed.weaknesses || [],
examples: parsed.examples || {},
suggestions: parsed.suggestions || []
};
} catch (error) {
return this.createErrorResult(category, error);
}
}
validateScore(score: any): number {
const numScore = Number(score);
if (isNaN(numScore) || numScore < 0 || numScore > 100) {
return 50; // Default neutral score
}
return Math.round(numScore);
}
createErrorResult(category: string, error: Error): CategoryResult {
return {
score: 0,
feedback: `Analysis failed for ${category}: ${error.message}`,
strengths: [],
weaknesses: ['Analysis could not be completed'],
examples: {},
suggestions: ['Please try re-running the analysis']
};
}
}Competitive Advantages Over AutoCrit
1. Speed Advantage
- AutoCrit: Several minutes to hours for analysis
- Our Approach: Sub-5-minute analysis with parallel processing
2. Depth of Analysis
- AutoCrit: Basic pattern recognition
- Our Approach: Contextual AI understanding with specific examples
3. Actionable Feedback
- AutoCrit: Identifies problems
- Our Approach: Identifies problems + provides specific improvement suggestions
4. Genre Intelligence
- AutoCrit: Basic genre comparison
- Our Approach: Deep genre-specific analysis with targeted feedback
5. Multi-Model Validation
- AutoCrit: Single analysis engine
- Our Approach: OpenAI + Claude validation for higher accuracy
Implementation Priority
Phase 1: Core Parity (Week 1-2)
- Pacing analysis with slow section detection
- Dialogue analysis with tag/adverb flagging
- Character development tracking
- Basic plot structure analysis
Phase 2: Advanced Features (Week 3-4)
- POV consistency checking
- World building analysis
- Strong writing evaluation
- Genre-specific refinements
Phase 3: Quality Enhancement (Week 5-6)
- Multi-model validation
- Confidence scoring
- Error handling and fallbacks
- Performance optimization
This AI-only approach will match AutoCrit’s analysis categories while providing faster, more detailed, and more actionable feedback for authors.