Automated consistency analysis

Definition

Automated consistency analysis compares established facts from the story — extracted from the manuscript and consolidated in a Narrative knowledge graph or bible — to detect contradictions: a character with a leg injury who runs in the next chapter, a shifting age, an object destroyed then reused, a place one hour away then described as three days' march. It handles factual and structural consistency, not literary quality or emotional coherence (a character may credibly change mind without triggering a flag).

Engines combine deterministic rules (if A precedes B, then X is impossible), graph queries, and sometimes LLMs with AI contextual memory for ambiguous cases. Each alert remains a hypothesis for the author to validate: deliberate ellipsis, unreliable narrator, or fuzzy magic rules can produce false positives.

What it's for

It frees cognitive bandwidth for meaning by automating fact checking. On a 100,000-word manuscript or multi-volume series, manual rereading of every detail becomes prohibitive; automated analysis targets risk zones so the author can focus revision effort. It complements — does not replace — human Beta reading and Editorial reading committee, which judge emotional impact and readability.

Integrated with a Narrative copilot (AI), it can run continuously (alert after each imported scene) or in batch before editorial submission. Paired with Narrative entity extraction, it improves as the graph sharpens. Treat alerts like margin notes from a sharp beta reader: worth reading, never worth obeying without your own judgment.

Variants by genre

• Fantasy / SF: world-rule checks — magic cost, technology range, map distances, species abilities.
• Mystery: chronological and logical checks — alibis, presence at crime scenes, clue consistency with the final solution.
• Romance: light relational checks — couple status, promises kept or broken; fewer physical rules, more factual emotional continuity.
• Literary fiction: minimal mode — gross factual contradictions only; respect ellipses and nonlinear structures.
• Long series: cross-volume analysis — Internal coherence recalls between books, detection of involuntary retcons.

Common pitfalls

• Blind trust: auto-correcting without verification — AI hallucinates or misses narrative intent.
• Stale codex or graph: useless or wrong analysis if the manuscript diverged from the bible.
• Confusing factual and emotional consistency: a change of heart is not an inconsistency; a character "too cold" after reconciliation may be intentional without factual contradiction.
• Rules too rigid: rejecting every time ellipsis or genre-assumed geographic jump.
• Ignoring repeated false positives: disabling the tool instead of refining schema or bible.
• Replacing human judgment: submitting because "the AI found nothing" without rereading.
• Expecting emotional diagnostics: the tool flags that a character was in Paris on Tuesday and London on Wednesday — not that a breakup scene feels unearned.

Examples

• A Song of Ice and Fire — fan communities documented minor timeline inconsistencies; Automated consistency analysis on a temporal graph would have targeted those zones before publication.
• The Girl with the Dragon Tattoo — procedural investigation: alibis, dates, digital assets; mystery is the genre where factual analysis adds the most value.
• The Time Traveler's Wife — explicit time-travel rules; analysis can verify each displacement respects stated constraints without judging scene quality.

Writing tips

• Keep a living rules log: every law of your world (magical, physical, social) deserves an entry with source chapter and first-appearance date. Automated analysis invents nothing — it verifies what you posed.
• Run a dedicated pass between the 1st and 2nd draft, then a second one 30 days before submission: two moments when inconsistencies surface most.
• Rank alerts by severity — an age that shifts by 3 years breaks credibility, a divergent sensory detail is minor. Never treat all alerts equally.
• Manually validate every alert: the tool flags, the author decides. Never rewrite blindly on the strength of a detection.

Learn more: Complete guide: novel coherence on 100,000+ words and How to maintain coherence on 100,000+ words.

Revision checklist

• Do Internal coherence and this notion reinforce each other, or does one contradict the other?
• Could a beta reader name Automated consistency analysis after reading without being told?
• Am I keeping a scene out of attachment that doesn't illustrate this notion?
• Did my last pass fix structure before style (see Line editing / polishing if needed)?
• Does the climax or pivot scene clearly show what Automated consistency analysis changes in the story?