Criteria-led reading
The manuscript is read through stable editorial criteria so feedback is not a random chat response or a matter of taste alone. The same core language lets authors compare reports across drafts.
Methodology
How RevisionGrade thinks about manuscripts.
RevisionGrade evaluates manuscripts through stable criteria, evidence-backed diagnosis, and revision restraint. The goal is not generic feedback. The goal is to identify what the manuscript asks the reader to believe, where that trust holds, and where it breaks.
Method standard
Stable criteria. Traceable evidence. Author-controlled revision.
The manuscript is not judged by vibes.
The report should explain why a finding matters.
Revision follows diagnosis; it does not replace it.
Reading principles
Four principles shape the evaluation.
The methodology keeps the product from becoming a generic critique engine. It reads through stable editorial dimensions, requires evidence, respects long-form behavior, and avoids unnecessary rewriting.
Evidence before verdict
Major findings should be traceable to the submitted pages and explained in terms the author can act on. A score without evidence is a verdict, not a diagnosis.
Long-form continuity
Novel-length work requires attention to recurrence, payoff, pacing over distance, character behavior, escalation, closure, and cumulative reader experience.
Revision restraint
A recommendation is not automatically an instruction to rewrite. Some passages should be repaired; others should be protected, preserved, clarified, or left alone.
Evaluation modes
Different manuscript lengths require different diagnostic promises.
RevisionGrade should not pretend a chapter can prove what only a full manuscript can show. Evaluation mode controls what the report can responsibly diagnose.
Under 25,000 words
Short-Form Evaluation
Core story diagnosis
Evaluates the submitted pages against the 13 story criteria only. Designed for openings, chapters, excerpts, short stories, and shorter works where full manuscript continuity cannot yet be judged.
25,000+ words
Long-Form Evaluation
Manuscript-scale readiness
Adds manuscript-scale analysis: continuity, recurrence, setup/payoff, pacing over distance, character behavior, structural readiness, and cumulative reader experience.
Complex long-form manuscripts
Long-Form Multi-Layer Evaluation
Deep architecture review
A deeper architecture audit for manuscripts that need layered evidence views, manuscript-scale continuity, proprietary repair governance, dialogue and speech protection, and deeper structural analysis where appropriate.
Thirteen Story Criteria
The core evaluation language stays stable.
The 13 criteria give every report a common vocabulary. Short-form evaluations use these criteria only; long-form and multi-layer evaluations build on them where manuscript length supports deeper claims.
Concept
Whether the central idea is clear, compelling, and strong enough to sustain the promised story experience.
Narrative Drive
Whether the story generates pressure, escalation, consequence, curiosity, and forward motion.
Character
Whether characters behave with coherence, agency, contradiction, development, and emotional credibility.
Voice
Whether the prose creates a distinct, controlled, and appropriate narrative presence without accidental flattening.
Scene Construction
Whether scenes are built around action, tension, change, stakes, and consequence rather than static explanation.
Dialogue
Whether speech reveals character, pressure, relationship, subtext, rhythm, and scene movement.
Theme
Whether the manuscript develops meaning through dramatic pressure instead of lecture, repetition, or abstraction.
Worldbuilding
Whether setting, context, culture, rules, history, and environment create a credible story world.
Pacing
Whether momentum, compression, expansion, turns, and recovery beats are proportioned to reader attention.
Prose Control
Whether the line-level writing is precise, intentional, readable, and aligned with the manuscript’s voice.
Tone
Whether the emotional and stylistic register remains coherent, deliberate, and appropriate to the material.
Narrative Closure
Whether the manuscript resolves, withholds, or complicates its central promises in a satisfying way.
Marketability
Whether the manuscript can be positioned clearly for a likely readership, category, shelf, or submission path.
Evidence model
A finding should move from observation to consequence.
The report should not merely name a weakness. It should explain what was observed, where it appears, how it affects the reader, and what kind of revision decision follows.
Observation
What the manuscript is doing on the page: repeated behavior, missing pressure, unstable voice, over-explanation, payoff drift, or a strong craft choice.
Evidence anchor
Where the observation appears: passage, chapter, scene, pattern, structural span, report section, or long-form layer.
Reader effect
Why the issue matters: confusion, loss of trust, reduced pressure, emotional flattening, promise drift, or increased engagement.
Revision implication
What kind of intervention is warranted: preserve, clarify, compress, restructure, escalate, defer, or repair in Revise.
What makes the diagnosis different
Useful feedback is specific enough to act on.
Generic critique tells the author to improve. RevisionGrade should identify the story behavior that causes the reader effect.
Generic feedback
“Increase tension.”
Evidence-backed diagnosis
The river scene states the emotional contradiction instead of dramatizing it; the reader receives explanation where hesitation, silence, or action would create pressure.
Generic feedback
“Make the protagonist more active.”
Evidence-backed diagnosis
Across the midpoint sequence, the protagonist observes consequences but rarely makes a choice that changes the scene’s direction, weakening perceived agency.
Generic feedback
“The pacing is slow.”
Evidence-backed diagnosis
Chapters 12–14 hold conversation without a consequence-bearing turn, creating a pressure plateau before the next major reversal.
Evaluate → Revise bridge
The method does not stop at diagnosis. It controls what happens next.
Evaluation identifies the weakness. Revise turns the weakness into a governed repair opportunity: evidence, diagnosis, options, voice risk, and explicit author decision.
01
Choose scope
The submitted word count and project type determine whether the report is short-form, long-form, or long-form multi-layer.
02
Read through criteria
The manuscript is evaluated through the 13 story criteria so the diagnostic language remains stable.
03
Anchor claims
Major conclusions should connect to manuscript evidence, confidence, severity, and reader effect.
04
Route revision
Findings can become governed repair opportunities in Revise, where the author chooses what changes.
Method boundaries
Responsible diagnosis requires responsible limits.
The methodology is strongest when it says what the system can diagnose and what it should not pretend to know.
Short-form evaluation does not claim full-manuscript continuity diagnosis.
Golden Spine/WAVE language belongs in long-form or multi-layer contexts, not short excerpts.
A score is not a publication guarantee or agent-interest prediction.
Marketability is positioning diagnosis, not a promise of sales or representation.
Revision recommendations remain proposals until the author decides.
External research should support context, not overrule the submitted manuscript.
Next step
From method to manuscript action.
Methodology explains how the system reads. Reliability explains why the author remains protected. Revise turns findings into controlled repair decisions instead of blind rewriting.