Narrative Consistency Checker

Narrative consistency checker

Long fiction is where continuity errors hide: a character’s eyes turn from blue to brown between chapters, an age drifts, a sibling becomes a cousin. This tool splits your manuscript into chapters, scans each one for explicit statements about your characters, and flags where a later chapter contradicts an earlier one. It runs entirely in your browser, so even an unpublished manuscript stays private.

How it works

The checker first detects chapter boundaries from “Chapter N” headings or divider lines. Within each chapter it runs pattern matching to capture attribute statements — phrases such as “her eyes were green,” “he was thirty-two,” or “his wife” — and records the character name, the attribute, the value, and the chapter it appeared in. After scanning, it groups every statement by character and attribute. Wherever the same character and attribute hold two different values, it reports a contradiction with both values and both chapter numbers, so you can jump straight to the conflicting passages.

The categories of errors this catches

Continuity errors cluster into a few well-known types in long fiction, and the tool is focused on the ones that pattern-matching can reliably surface:

Physical attribute drift is the most common and most embarrassing continuity error because readers remember physical details vividly. Eye colour, hair colour, height described as “tall” in chapter two and “average” in chapter twelve, a scar that appears and disappears — these are the errors that generate reader complaints and editorial notes.

Age inconsistency arises when a character’s stated age in an early chapter does not align with a later statement, especially in stories where months or years of in-world time pass between scenes. The tool captures explicit age statements (“she was forty-two,” “the twenty-year-old”) and flags when the same name is later given a different number.

Relationship contradictions — a character introduced as a colleague who later appears as a childhood friend, a sibling who becomes a half-sibling — are caught when the same possessive relationship phrase changes between chapters.

What the tool cannot catch

Being honest about the limits is as useful as knowing what works. The pattern-matching approach does not catch:

• Implied contradictions — a character who is established as fearless in chapter one but acts cowardly in chapter ten without any explicit statement of either fact
• Timeline inconsistencies — travel times, dates, seasons that add up wrong without any single sentence contradicting another
• Pronoun-only references — a character referred to only as “she” cannot be identified without a name anchor
• Plot logic gaps — a character knowing information they were not present to learn

For these, the tool is a complement to human continuity reading, not a replacement.

Formatting your manuscript for best results

The quality of the output depends significantly on how clearly your chapter breaks are marked:

• Chapter 1 or Chapter One on its own line — detected reliably
• --- or *** divider lines — detected as section breaks
• No dividers at all — the entire manuscript is treated as one section; contradictions are still found but without chapter-number attribution

Character names must appear in the same form throughout for the matcher to group statements correctly. If you write “Elizabeth” in early chapters and “Beth” later, the tool sees two different characters. A global search-and-replace to standardise names before pasting will produce much cleaner results.

Tips and notes

• Name characters consistently. The matcher keys on names, so “Beth” and “Elizabeth” are treated as two people unless you make them match.
• It is a first pass. Explicit, restated facts are caught; implied or pronoun-only details are not. Always follow with a human read.
• Mark your chapters. Clear “Chapter N” headings give the most useful, chapter-numbered output.
• Private by design. Nothing leaves the browser, which matters for unpublished work.