Continuity Checker for Fiction Writers

A continuity checker reads a manuscript, extracts the facts it establishes — character details, locations, timeline events, world rules — and flags any new sentence that contradicts them. In PlotLens the extraction is automatic; the flagging is on demand: upload a manuscript in .docx, .pdf, .epub, or .fountain and get a story bible pulled from your own prose, with every new upload checked against the established facts inferred from it. The full AI continuity analysis — with sentence-level citations — runs on the chapters you ask it to check, as a batch report, live in the editor, or as a spot check. Nothing is generated; every flag cites the passage that established the fact.

Spell checkers catch typos. Grammar tools catch syntax errors. But until PlotLens, nothing caught the moment your protagonist’s sister changed names between chapters. PlotLens validates narrative continuity — the layer of consistency that makes fiction feel real.

The three kinds of continuity errors that ship

The errors that show up in published fiction are usually one of three shapes, and they’re all invisible to spell- and grammar-checkers because they’re not language errors — they’re fact errors.

The first is detail drift. A character’s eye color, the spelling of a town’s name, a scar described on the left cheek in chapter four and the right cheek in chapter eleven, a horse named in dialogue but never again. These slip past every human reader who isn’t holding a spreadsheet because no human can hold the whole text in working memory.

The second is timeline contradiction. The story says it’s been three weeks since a key event, but the actual chronology of the chapters that follow only accounts for ten days. A character travels from city A to city B in a day in book one and three days in book two without explanation. Two scenes that were supposed to be simultaneous reference each other as having already happened.

The third is state contradiction — usually about what a character knows. A character references a secret they haven’t been told yet. A character claims not to know something the story already showed them learning. A character forgets a fact that’s important to a later scene. These are the hardest to catch by hand because they require simulating each character’s epistemic state at every point in the narrative.

PlotLens is built to catch all three. Detail drift falls out of entity extraction with provenance. Timeline contradiction falls out of the time-indexed Story Bible. State contradiction is what the belief-vs-fact model is for.

Your Story Bible is the set of facts your story has established — what readers call canon.

How PlotLens actually validates continuity

The flow is simple to describe and the result is what matters: you write a chapter, PlotLens compares every sentence in it against your Story Bible and any sentence that contradicts those facts shows up as an inline annotation with a citation back to the entry it conflicts with. Click the annotation and you see both passages side by side — the new sentence and the earlier sentence that established the fact. Decide what to do: fix the new sentence, accept the new state as an established fact (a retcon), or mark the contradiction as deliberate (a character lying, a narrator being unreliable).

What’s behind the simple flow:

• The Story Bible is derived from your manuscript. Nothing in the bible was typed in by hand. The checker is comparing your new prose against extractions from your earlier prose, with sentence-level provenance on every fact.
• Entity resolution handles the messy cases. “Detective Morrow,” “Morrow,” “Anna,” and “Anna Morrow” resolve to one character without you teaching the system anything. Aliases, nicknames, titles, and the way characters refer to each other in dialogue all collapse to the right identity.
• The validator distinguishes evidence types. A fact established in narration is treated differently from a claim made in a character’s dialogue (which might be true, might be a lie, might be the character’s mistaken belief). The system doesn’t flag a character lying as a continuity error; it flags the narrator describing a fact that contradicts an earlier narrated fact.
• Timeline math is explicit. Time references — “three weeks later,” “the next morning,” “that Wednesday” — get reconciled against the rest of the timeline. Contradictions show up as flagged spans, not as guesses.

Where it fits in a working writer’s day

A novelist working through a third draft uses PlotLens as the pass between “the writing is done” and “send to the editor.” Run extraction once. Open chapter one in your editor. Read through with the validator active. Every flag becomes a decision: fix, retcon, deliberate. By the end of the read-through, the manuscript is internally consistent and the bible reflects the final state.

A screenwriter on a feature uses the same pattern at the table-read stage. Validation runs against every scene in the script, with screenplay-aware parsing (scene headings, character cues, action lines, dialogue all treated as distinct evidence types). Continuity errors get caught before they cost reshoots.

A writers’ room on a serialized show runs the checker after every script draft. The shared Story Bible catches the writer in episode 405 who’s about to contradict a reveal from episode 207. The room debates whether to retcon or fix; the bible records the decision; future scripts are checked against the new state.

If you’re tired of finding the eye-color drift in the reader review, the free tier covers a single project so you can run the validator on a manuscript you’ve already finished. Compare PlotLens to Sudowrite for the analysis-vs-generation framing or to Novelcrafter for the closest tool that pitches a similar workflow. The novelists surface, screenwriters surface, and game-studios surface each describe the validation flow in the medium-specific shape it takes for that audience.