AI Manuscript Analysis, Book by Book: Inside StoryHelm’s 19-Analysis Full Read
StoryHelm’s Full read runs nineteen analyses over each book you import, from its structure and scenes to its timeline, foreshadowing and voice, and checks the book against its own canon. It reads one book at a time; a separate cross-book check you start compares the books.
The continuity review is the Full read: nineteen analyses that read one book of your series at a time against that book's own Canon, run in dependency-based waves, and hand back a severity-ranked verdict tied, in most cases, to the chapter and scene where each issue lives. A separate cross-book check names the books where a character's recorded details disagree, or where two fully read books give the same event a different year, month, season or length. StoryHelm only ever reads your prose; you write every word.
Most analysis tools read one book in isolation. That is fine for a standalone novel, but a series works differently. The slip that matters most is rarely inside book three; it sits between book one and book three, separated by a couple hundred thousand words and a year and a half of your life. To catch a drifted eye color before a reader does, something has to set what each book records side by side.
What follows is what the nineteen analyses look at, the order they run in, and what lands on your desk at the end.
The principle: read each book, then compare the books
A single-book tool builds a fresh picture of your world every time it opens a file. Book three knows nothing about book one. So when your hero's eye color shifts, when a character who died in book two answers the door in book four, or when a sword you established as one-of-a-kind shows up in two hands at once, a single-book read simply cannot see it. There is no shared memory to compare against.
StoryHelm reads one book at a time too, but it keeps what each read finds. Every book gets its own Canon, a structured record of the characters, places, factions, items, and events its read found, and the Series Atlas gathers every book's Canon in one place. A separate cross-book check, which you start from your series' Checks page, compares what each book records about characters whose names match or are spelled almost the same: a physical description, a scar, and, when both books hold enough of that character's dialogue, the way they speak. So the gray eyes in book one's record are still there to set beside the brown that book five's read records. The check also flags the same event given a different year, month, season or length in two fully read books. The dead man at the door and the doubled sword are not compared across books; those are still yours to check. Watson’s wandering wound shows what a cross-book finding looks like. The AI Beta Reader runs separately, after a read, and this guide says what kind of feedback the read is.
One book in isolation
- Fresh, empty memory per book
- Cannot compare book one to book four
- Misses drift that crosses installments
- Forgets each book once it closes
Each book, then the whole series
- Every book's Canon in one Series Atlas
- A cross-book check on character details and event dates
- Flags a changed appearance between distant installments
- Foreshadowing tracked from setup to payoff
The contradiction that matters most is rarely inside one book. It is between books.
The 19 analyses, grouped by theme
The review runs nineteen distinct analyses. Each is a focused pass with a single job, and each writes into the same shared picture so the next pass can build on it. Here they are, grouped by what they examine.
Why nineteen and not twenty? The review is built from twenty passes, but one of them, Context Bundles, which pre-selects the characters, places, and objects each scene involves, is switched off by default. A Full read runs the other nineteen.
Shape of the book
The Structure Analyzer maps chapter boundaries across the full text; the Outline Extractor turns that into acts and a clean scene-by-scene skeleton you can actually read.
Who and what is in here
The Manuscript Extractor pulls the characters, places, factions, items, and events it finds into that book's Canon, so the rest of the read has one record to check against.
The pulse of each scene
Plot DNA pulls the book's conflicts, character arcs, storylines, and themes, while the scene scores taken at import chart intensity, stakes, and the filler ratio; the Scene Content pass gives several of the passes after it real scene text to work from.
Does it hold together
The Narrative Analyzer reads pacing, tension, and point of view scene by scene, and the Coherence Guardian checks the book against its own Canon for timeline, trait, place, object, and who-knows-what contradictions, the heart of the continuity read.
Setups and voices
The Foreshadowing pass pairs setups with their payoffs within each book, and lets you flag setups that pay off in a later book so they are not counted as unresolved; the Dialogue extractor captures who says what, the raw material for the voice checks.
Reconciling the record
The Cross-Validator checks the Canon, the outline, and the scene record against each other; the Timeline pass puts the book's events in story order and flags a character in two places at once or events out of order.
How it reads on the page
Voice Consistency tracks whether a character still sounds like themselves chapter to chapter; the Style Coach flags overused words and clunky sentence patterns, and the Sensory Audit checks which of the five senses each scene leans on.
The laws of your world
The World Rules extractor captures the rules of magic, tech, and society; the Validator and the Optimizer then score how complete the read is, without re-running any pass.
The verdict, assembled
Narrative Synthesis pulls the separate findings into a single read on the story; Diagnostic Synthesis assembles the severity-ranked report you actually open, with chapter and scene references.
This continuity review is the Full read, which comes with the Author, Series, and Studio plans; the Quick scan on the Writer plan runs 7 of these passes and skips the Coherence Guardian and the Timeline pass. The Full read answers the first question behind every long series: does each book still agree with itself?
Why it runs in waves
The analyses are not independent. A timeline check is meaningless until events have been extracted; a foreshadowing trace needs scenes in hand before it can follow a promise to its payoff. So the review runs in dependency-based waves: each wave produces facts the next wave depends on.
-
Lay the foundation
Structure first, then the outline, the manuscript extraction, and Plot DNA. The skeleton and the cast have to exist before anything can be checked against them.
-
Read the scenes and the story
Scene content feeds the Narrative Analyzer, foreshadowing, and dialogue, while the Coherence Guardian works from the Canon, Plot DNA, and the outline. This is where most of the book's contradictions surface.
-
Reconcile and refine
Cross-validation, timeline ordering, and the voice, style, and sensory passes reconcile and deepen the picture, then the world rules pass runs and the Validator scores how complete the read is.
-
Synthesize the verdict
Narrative and diagnostic synthesis assemble everything into one severity-ranked report. Just before them, the Optimizer notes whether the read passed its own quality bar; it does not re-run any pass.
Because the work is ordered by dependency rather than crammed into a single prompt, each pass examines one thing closely with everything found before it already in hand.
What you get out: the series verdict
The output is not a wall of raw data. It is a severity-ranked verdict on each book, anchored in most cases to the chapter and scene where each issue lives, plus, when you run a cross-book check, a list that names the books that disagree.
You can take that verdict deeper through the craft lenses. The one-time $99 Manuscript Audit, arranged with the StoryHelm team, runs this same Full read on one manuscript, then an AI Beta Reader pass with its reader's letter, and hands back a prioritized report. On the Series tier, your 1,500 monthly credits cover about three Full reads of a standard-length novel, and a re-read that skips unchanged chapters costs about a quarter as much.
StoryHelm keeps your whole series in one place. You write and edit your books in its built-in Scene Editor, or import finished manuscripts written elsewhere, with your Canon and Series Atlas alongside them. From there the continuity review, one workflow in a 61-pass engine powered by Claude, reads each book against that book's Canon and returns a severity-ranked verdict with chapter and scene references. A cross-book check names the books where a character's recorded details disagree, or where two fully read books give the same event a different year, month, season or length. You write every word. StoryHelm makes sure the words agree with each other.
One workspace, not a separate scan
People get one thing wrong about this review: it is not a tool you run a finished book through once and then put away. It lives inside the workspace where you write. Draft a new chapter in the Scene Editor, and each save runs quick checks on that scene against the book's Canon, such as a name spelled almost like one on record; you see them when you open the scene or press Rescan. The full nineteen-analysis review is the deep, deliberate read, and you re-run it on your plan's credits as the series grows. Same Canon, one place.
Books written in another tool come in as Word (.docx), PDF, RTF, EPUB, plain text, or Markdown files; save an older .doc as .docx first. Each book's read fills its own Canon. The cross-book check runs only when you start one from your series' Checks page: you choose at least two books with a completed read and see the credit cost before you confirm. Adding or removing a book can leave a reminder on your series overview suggesting a check; finishing a book's read does not start one.
FAQ
No. It reads, checks, and analyzes your manuscript, and it never drafts or rewrites your prose. You write every word. Two tools do produce text, and say so: Ship Dock drafts pitch material about your book, such as a query letter or synopsis, for you to review and edit, and Translation Studio produces machine translations of your finished book.
Nineteen distinct analyses, each with a single job, running on one book in dependency-based waves so later passes can build on earlier findings. They share that book's Canon and assemble into one report.
One at a time. Each book's read fills that book's own Canon, and the Series Atlas gathers every book's Canon in one place. A separate cross-book check, which you start from your series' Checks page, compares what each book records about characters whose names match, such as a physical description or a scar, and flags a noticeable change in how a character speaks when both books hold enough of their dialogue. It also flags the same event given a different year, month, season or length in two fully read books. A finding names the two books that disagree and shows what each one records.
Because the analyses depend on each other. A timeline check needs extracted events; a foreshadowing trace needs scenes. Running in waves lets each pass work with everything found before it.
Yes. StoryHelm’s AI reads and analyzes your manuscript and reports what it noticed. It never rewrites your prose; you decide every change. Two disclosed features produce text that isn’t your manuscript: Ship Dock drafts pitch material about the book for you to review and edit, and Translation Studio produces machine-generated translated editions of a finished book.
StoryHelm’s does: it runs on Claude, built by Anthropic. Your manuscript isn’t used to train models; Anthropic keeps API inputs and outputs for a limited window before deleting them, as the privacy page explains.
Keep reading
See the verdict your readers would have caught.
Bring in book one or all six, and watch a Full read (Author plan and up) run nineteen analyses on each book against its own Canon. Then start a cross-book check that hands back the books where a character's recorded details stop agreeing, or where two fully read books give the same event a different year, month, season or length. Find the slip between book one and book five now, not in a one-star review after the box set ships.
Get the free Self-Audit