Belief vs Fact

Most continuity checks ask “is this true in the story?”. Belief vs Fact also asks “does this character actually know it yet?” — so a kept secret, a lie a character tells, or a fact someone hasn’t been told doesn’t trip a false-positive contradiction.

What this does

PlotLens tracks a separate state for each character: what they believe as opposed to what is true in the story world. When you run validation in POV mode, the validator checks new writing against that character’s known beliefs, and surfaces two extra kinds of finding — knowledge violations (a character acting on a fact they shouldn’t know yet) and forgotten reveals (a character ignoring something the story already told them). You can mark any of these as deliberate and they stop re-appearing on the next run.

When to use POV-scoped validation vs objective mode

PlotLens has two validation modes. They check against different things.

Objective mode is the default. It checks the new text against your project’s canon — the established facts, relationships, and rules. Use it when you want to ask “is this consistent with the world?” — for example, a worldbuilding pass, a continuity edit, or a draft sweep where you want every contradiction surfaced regardless of POV.

POV-scoped mode is the new one. You pick a character on the validation form, and the validator checks the text against that character’s known beliefs instead of treating every fact as universally known. Use it when:

• You’re writing in close third or first person from a single POV and want feedback that respects what the narrator actually knows.
• A character is meant to be unaware of something — a kept secret, a hidden identity, a twist that hasn’t landed yet.
• A character is meant to believe something false — a lie they’ve been told, a misdirect they’ve fallen for, a misremembered detail.
• You’re auditing a scene for forgotten reveals: a character was told the killer’s identity in chapter 3 but writes as if they don’t know in chapter 8.

If the picker is empty on the validation form, your project doesn’t have a POV mode opt-in yet — fall back to objective mode for the run. (See Plan availability and flag state below.)

How to read a Knowledge state result

When you run validation in POV mode and the validator finds anything new, the results page shows a dedicated Knowledge state section in addition to the usual errors, warnings, and info findings. The section has two buckets:

• Knowledge violations — places the new text shows a character acting on a fact they wouldn’t know at this point in the story.
• Forgotten reveals — places the new text shows a character missing or contradicting a fact they have already been told earlier in the story.

Each finding includes:

• The character the finding is scoped to (name plus a small avatar chip).
• A short claim describing the knowledge being checked (e.g. “knows the queen is alive”).
• A transmitting scene citation — the earlier passage where the character either learned the fact (for forgotten reveals) or where the fact appears in the world without being shared with the character (for violations). If no transmitting scene was found, the section says so.
• A violating scene citation — the passage in the text under validation that contradicts what the character should know.
• Reason-code chips that explain why the validator flagged this finding (for example, the kind of inference path it followed). The chips are short uppercase tokens; see the reason-code reference for what each one means.

Treat the citations the same way you treat any other PlotLens finding — they are how the validator shows its work. If the citation doesn’t ring true, the finding is probably wrong; if it does, the finding is probably real.

Marking a tension as deliberate

Some belief tensions are the point — a character lying, an ironic gap the reader is meant to see, a deliberate forgotten reveal that will pay off later. You don’t want those re-emitting on every validation run.

To mark a finding as deliberate:

1. Open the Knowledge state section on the validation results.
2. Find the knowledge violation or forgotten reveal you want to suppress.
3. Click Mark as deliberate on the card.
4. (Optional) Add a short note — for example, “intentional dramatic irony” or “sets up the reveal in chapter 14”. The note is for your future self; PlotLens uses it as documentation, not as a signal.
5. Click Confirm.

The card collapses into a “Marked as deliberate” pill and the finding stops appearing on subsequent runs of the same scene against the same character. The mark is sticky per character + fact, so if a different character has the same kind of tension you’ll still see that one.

If the underlying scene materially changes — for example, you rewrite the violating passage so the citation is no longer accurate — the finding can re-emit. That’s by design: a meaningful edit may have removed the intentional gap you were preserving.

What the belief statuses mean

PlotLens stores four possible beliefs per character per fact. You’ll see them on a character’s belief-vs-fact tab in the entity detail page, and they’re the values the validator compares against in POV mode.

Status	Plain English
BELIEVES_TRUE	The character holds this fact as true. They’ve been told, they’ve witnessed it, or they’ve worked it out — and the story has them acting on it.
BELIEVES_FALSE	The character holds the opposite as true. They’ve been lied to, they’ve drawn the wrong conclusion, or they’re operating on outdated information.
UNAWARE	The character has not encountered this fact yet — neither the true version nor a false one. The character should not write or act as if it’s known to them.
SUSPECTS	The character has a hunch but no confirmation. They might allude to the fact, hedge, or test it — but they shouldn’t treat it as settled.

The validator uses these statuses to decide whether a passage is a contradiction for that character. A passage where the character treats an UNAWARE fact as known is a knowledge violation. A passage where a character contradicts a fact whose status is BELIEVES_TRUE (and they were told it on-page) is a forgotten reveal. A passage that aligns with BELIEVES_FALSE is not a contradiction — the character is meant to be wrong here.

If a character’s belief is autopopulated to the wrong status, you can edit it on their entity page. Manual edits win over autopopulated ones.

How this relates to canon rules

Belief vs Fact does not replace canon rules — it sits next to them. Canon rules answer “what is true in this story?”. Belief vs Fact answers “and who knows it?”. The two systems share the same underlying epistemic metadata: a canon rule marked as unreliable_pov, rumor, or speculation is already softened in the validator’s eyes, regardless of whether you’re running in POV mode or objective mode. See Canon rules for how to set rule-level epistemic status, and Validation for how findings of every kind are surfaced.

Plan availability and flag state

Belief vs Fact is available as part of validation on Plus and above. Free and Lite plans do not include validation, so they do not see the POV picker or the Knowledge state section.

Plan	Belief vs Fact
Free	Not available
Lite	Not available
Plus	Available — counts against your monthly validation allotment
Pro	Available — unlimited (fair use)
Small Team / Studio / Production	Available — per-seat, unlimited (fair use)
Enterprise	Available — custom

A POV-scoped validation counts as one validation against your monthly allotment, the same as an objective-mode run. See Billing & plans for the full plan matrix.

Flag state. Belief vs Fact is rolling out gradually. If the POV picker is missing from your validation form even on a paid plan, the feature is not yet enabled for your workspace — you’ll see the SPA fall back to objective-mode-only validation, with no Knowledge state section in the results. Once the feature is enabled on your account, you’ll see it across all of your projects. Nothing about your existing canon or validation history changes.

Limits and edge cases

• No POV characters in the picker. The picker only lists character entities for the current project. If you haven’t uploaded a document or added a character yet, the picker is empty — pick objective mode for now.
• Character with no autopopulated beliefs. A newly-added character may have an empty belief profile until the autopopulate job runs. POV-mode validation against an empty profile will return zero Knowledge state findings — not because the scene is clean, but because there’s nothing to compare against. Wait for autopopulate to complete, or add a few beliefs by hand on the entity page.
• Forgotten reveal with no transmitting scene. Occasionally the validator detects a forgotten reveal but can’t surface the exact transmitting passage. The card shows (no transmitting scene found) in place of a citation. The finding is still valid — the underlying belief is on the character’s profile — but you’ll need to track down the source scene yourself.
• Mark-as-deliberate revival. If you rewrite the violating scene materially, the deliberate mark can lapse and the finding can re-emit. Re-mark it if it’s still intentional.
• Objective vs POV disagreement. A passage can be a contradiction in objective mode but not in POV mode (a character knowingly lying), or vice versa (a character acting on a secret they don’t yet know). Picking the right mode for what you’re auditing is the whole point — don’t expect the two modes to agree on every finding.

Common pitfalls

• Running POV mode when you wanted a worldbuilding check. POV mode hides contradictions a character isn’t aware of. If you’re sanity-checking world facts, leave the POV picker empty and run in objective mode.
• Mass-marking findings as deliberate. Each deliberate mark suppresses the finding for that character + fact. Bulk-marking everything without reading the citations defeats the point — you’ll lose real signal alongside the noise.
• Forgetting to update beliefs after a major rewrite. If you rework the chapter where a character learns a key fact, their belief profile may now be wrong. Check the character’s belief-vs-fact tab on the entity page after big rewrites.
• Confusing BELIEVES_FALSE with a contradiction. A passage that lines up with BELIEVES_FALSE is correct — the character is meant to be wrong. Don’t fix it.

Related

• Canon rules — what counts as a fact, and how epistemic status (fact, rumor, lie, dream, unreliable_pov, speculation, red_herring, retcon) softens or excludes rules from validation
• Validation — the three validation modes (batch, real-time, spot-check) and how conflicts are surfaced
• Reason codes — the full reference for the reason-code chips on knowledge findings
• Entity extraction — where characters come from in the first place
• Billing & plans — validation allotments and what each plan includes