What did my character know in chapter N?

You are planning chapter forty, and Kai has to warn Mia about the King, and you stop: does he actually know this yet? He overheard something in the throne room, or was that the draft you cut? If you cannot answer without rereading, every scene built on the answer is built on sand. Knowledge continuity is the discipline of making "who knew what, and when" a lookup instead of a guess.

The failure is quiet. Unlike the shattered sword that reappears, nothing on the page immediately contradicts a character who knows too much. The scene reads fine, ships, and then the reveal lands flat because the reader never saw the character earn the information. In anything mystery-shaped this reads as cheating, and it is among the continuity errors readers are least willing to forgive (see the seven continuity errors).

Keep a knowledge ledger

For every secret or significant fact in the story, keep one small record. The ledger has a line per fact, and for each person on that line:

• The knower, with the chapter and scene where they learned it on the page.
• The mechanism: told, overheard, witnessed, deduced. A secret deduced from two clues plays very differently from one delivered in exposition, and the mechanism decides how a later confrontation can be staged.
• Who must not know yet: as canonical as who does. This is the engine of dramatic irony, and losing track of it is how spoilers leak from author to character.
• Where it was established: the chapter and scene, so you can reread the exact passage instead of trusting your memory of it.

Like every other kind of character state, knowledge changes over time, so the ledger keeps history instead of holding only a current snapshot. "Mia knows the King is a dragon, since chapter 30, Kai told her after the duel" is a fact with a birthday: it was false in chapter twenty-nine and true afterward. The rule is the same date-it-don't-overwrite-it rule as character state tracking.

The three questions any tracker must answer

Knowledge is one of three state threads that break together. A tracker is working when, for any chapter you name, it can answer three questions without a reread:

1. What did this character know at that chapter?
2. Where was this character at that chapter?
3. Was this character injured, or otherwise changed, before that chapter?

If it cannot, the record was overwritten somewhere. The most common overwrite is the innocent-looking "current knowledge" field on a profile: it answers who knows now, and erases the answer to who knew then.

Evidence, or it is a theory

An entry with a source is worth several without one. Pin each piece of learning to its chapter, its scene, and ideally the passage itself, for three reasons:

• You can check yourself. What you remember you wrote drifts toward what you meant to write. The passage is the only authority.
• Revision gets faster. When a reveal feels weak in draft two, the evidence tells you in seconds whether the learning scene exists, or whether you skipped writing it.
• Disputes end. A knowledge claim that cannot point at a page is not an established fact; it is a theory you are carrying. Canon separates established from inferred for exactly this reason, and knowledge deserves the same rigor.

Check the ledger before every reveal

Make it a ritual before any confrontation, reveal or meeting scene:

1. List everyone who will be in the scene.
2. For each secret at stake, check who in the room knows it, how, and since when.
3. Check the reader's knowledge separately from every character's. Dramatic irony is reader-knows, character-does-not; surprise is neither-knows. Both collapse when you lose the distinction.
4. If someone must know something the page never showed, you have three honest options: write the learning scene, move the reveal later, or let another character deliver the news on the page now.

Off-page learning costs trust

You can grant knowledge off page when it is trivial: directions, public events, things everyone in town knows. Significant knowledge granted off page should be rare, and when it happens, acknowledge the gap on the page ("she had spent the week in the archives") so the reader sees that you know they would ask. A plot-critical secret should never arrive off page. The ledger makes this visible: an entry whose mechanism reads "somewhere before chapter forty" is a hole with a date on it.

How Kanonara approaches this

Secrets and what each character knows are structured records with history in Kanonara, part of each character's state over time. When a chapter is analyzed, a character acting on information no earlier scene established is flagged as a question, with the evidence for what the page does and does not show, and proposed knowledge updates arrive with the passage they came from. Nothing becomes canon until you approve it. See how Kanonara works.