A forensic investigation of your novel
The popular idea of an AI editor is a machine that rewrites your sentences. It reads a paragraph, proposes a better verb, smooths a clause. Useful, in its way, and almost entirely beside the point once a novel is finished.
Novel Maestro does something else. It treats a finished manuscript as a case to be investigated: a hundred thousand words of evidence, examined as one system rather than read a chapter at a time. Every character arc traced from first appearance to last. Every setup held against the payoff it did or didn’t get. Every scene-to-scene join tested for whether one thing genuinely causes the next. Every fact cross-referenced against every other fact in the book.
This matters because of what actually goes wrong in a finished novel. It is almost never the sentences. It is a thread that dies quietly in chapter nine and is never mentioned again. A payoff that arrives two hundred pages after the reader stopped waiting for it. A protagonist who, somewhere in the middle third, stops making decisions and starts being moved around by events. These faults are distributed across the whole book, which is exactly why reading it again won’t find them — and why you, of all readers, are the least likely to. You have read it too many times. The gaps have closed over in your memory, and the book you see is the one you meant to write.
How an investigation differs from an opinion
An opinion about a novel is a reaction. An investigation is a procedure, and it leaves a record.
Every finding Maestro returns is tied to the place it came from: the scene, the line, the quotation. When it reports a contradiction it names both of the conflicting scenes, rather than telling you that something, somewhere, doesn’t add up. When it scores a character’s arc it shows the evidence behind each score, and lets you overrule it where it has read you wrong. When it says a thread goes dormant it can tell you for how many scenes.
Nothing is taken on trust — including your own account of the book. What you meant is recorded separately from what is on the page, so the two can be compared rather than confused.
The findings then have to become work. They are sequenced into a revision programme that puts structure before scene and scene before line, so you are not polishing a chapter you will later cut, and it marks what is already working so that a rewrite doesn’t cost you the best thing in the book.
About the engine
Large language models do the reading. That is the engine, and it is worth being plain about it — but it is not the interesting part, any more than a laboratory is interesting for owning a centrifuge.
What matters is what gets examined, how rigorously, and what comes back. Maestro does not write your prose and will not offer to. It reads, it tests, it finds, and it hands you the findings. The book stays yours, which was always the point.
What an investigation turns up
50 things you can learn about a manuscript you thought you knew. Not a menu of features — findings, in the form you would actually receive them.
The shape of the book
How the novel moves across its whole length, rather than how any one chapter reads.
- Whether your novel accelerates, holds steady, decelerates or moves erratically — named, not guessed at.
- Where the energy drops far enough that a reader is likely to stop, marked on a chart of the whole book.
- Whether you have a sagging middle, a rushed climax or a flat opening, each located where it occurs.
- Which of six dimensions is dragging any given stretch: tension, conflict, stakes, agency, revelation or emotional intensity.
- How your word count sits against the band your genre expects, and what that implies for submission.
Cause and effect
Whether the book is a chain of consequences or a sequence of events that merely follow one another.
- Every scene-to-scene transition in the book marked causal or merely sequential.
- Your causal ratio — the proportion of the novel where one scene makes the next one happen.
- The exact runs where the story drifts into “and then”, and where it turns on coincidence rather than consequence.
Characters
Who is on the page, for how long, and whether they change.
- A presence timeline for every character, scene by scene, from first appearance to last.
- How long each character disappears for, and whether a reader would have lost them by the time they return.
- Which characters overlap in function closely enough that one of them is doing no distinct work.
- Each character's transformation tracked across eight dimensions, scene by scene.
- Whether an arc matches the shape its type implies or flattens against it — and exactly where it stalls or reverses.
- Your protagonist's want set against their need, and whether the gap between them is doing dramatic work.
Who is driving
The difference between a protagonist who acts and one who is acted upon.
- The share of scenes your protagonist actually drives, as a figure.
- The stretches where characters stop making decisions and start being moved by events.
- Who decides in each scene — and a warning when, across a long run, it isn't your protagonist.
- Where a character takes the easy way out of a choice the story had set up to be hard.
Scene economy
Which scenes move the story, and which only tell the reader something.
- One row per scene: who decides, what changes, what the reader learns.
- Which scenes change nothing and only deliver information — visible as empty cells, not as an opinion.
- What proportion of your scenes actually move the story.
- The stretches where the book stalls for more than a scene at a time.
- Which scenes could be compressed, combined or cut, named individually.
Threads and subplots
Everything the book has started, and what became of it.
- Every narrative thread identified and tracked — plot, relationship, thematic and subplot.
- Where each thread starts, which scenes it lives in, and where it goes quiet.
- The points at which a reader is likely to have forgotten a thread exists.
- What share of your scenes each storyline occupies, against a benchmark for main plot and subplot.
- Whether your main plot still dominates, or a subplot has quietly taken the book over.
Setups and payoffs
Every promise the book makes to the reader, matched against what it delivers.
- Every setup, planted detail and piece of foreshadowing found in the manuscript, listed.
- Each one's status: resolved, still hanging, paid off late, or forgotten entirely.
- How long the reader is asked to wait for each payoff.
- The promises your opening pages make, and whether the ending keeps them.
Continuity and world logic
Facts cross-referenced across a hundred thousand words, which is not a thing a person can do by reading.
- Story facts cross-referenced across the whole manuscript, and the contradictions between them.
- For each contradiction, both conflicting scenes named — not a flag that something, somewhere, is wrong.
- A confidence level on every finding, so a certain error reads differently from a possible one.
- Timeline slips: dates, ages, seasons and durations that don't reconcile.
- Two characters named closely enough to be confused — including when you have already acknowledged it in the text.
Genre and expectation
What readers of your kind of book arrive expecting, checked against what they get.
- A genre fulfilment score, with the kept, bent and broken tally behind it.
- Which scenes your genre expects that your manuscript doesn't contain.
- For each convention, the common version and how you might make it yours, with published comparables.
- Somewhere to log what you are breaking deliberately, so a choice is recorded as a choice rather than flagged again as a fault.
Theme and pattern
The material that recurs, and whether it accumulates into anything.
- Every recurring motif found — image, symbol, emotional pattern or thematic echo — and how often each returns.
- Whether each one is consistent, emerging, fading, or appears only once.
- The stretches where your theme goes quiet for long enough to disappear.
- Scenes where the prose tone contradicts the dramatic turn: the writing warm where the moment is a loss.
What a reader would say
The investigation ends where every book ends — with someone reading it.
- Six named readers on your manuscript: an avid reader of your genre, a literary agent, a commissioning editor, a character-driven reader, a first-time reader and a pacing hawk.
- Whether the agent would pass, ask for a revise-and-resubmit, or request the full manuscript.
- Exactly where a pacing-driven reader started to skim, and where a first-time reader lost track of who's who.
What comes of it
Findings are only useful if they become work you can actually do.
- Revision priorities sequenced structure before scene before line, so you don't polish something you will later cut.
- What is working that a rewrite could damage, named explicitly so that you protect it.
You cannot investigate your own novel. You are too close to it, and you have read it too often to see it.