Counted, not guessed.

Manuscript Analysis

Manuscript Analysis reads a finished draft and tells you what can be proven about it. The free layer is arithmetic over your own words: the statistics of the book, the things that would embarrass it on a store page, an estimate of where a shopper's free sample ends, and candidates where two numbers or two spellings of a name disagree. One paid reading goes further and looks for exactly two things. Nothing here rewrites a word of your book.

Analyze your manuscript See pricing

700+

books generated

600+

writers on the platform

book languages

DOCX · EPUB · PDF

KDP-ready exports

Live platform totals, verified 2026-07-17

Why it works

Everything countable is free

The statistics, the publish readiness check, the estimated sample cutoff and the number and name candidates cost nothing and run no model. Charging for arithmetic would be selling a calculator.

Every claim quotes your own words

A finding is a count, a position, or two passages placed side by side. You can check any of it in a second, which is why nothing here is called an error.

It never rewrites your book

Analysis reports and stops. What changes and what stays is your decision, and no button on these screens edits a sentence of yours.

How it works

1
Bring your draft in
Import a manuscript into Writer's Workshop, or open a book you wrote here. Analysis reads what is already there.
2
Read the free layer
Statistics, publish readiness, the estimated sample cutoff, and the number and name candidates. No credits and no waiting on a model.
3
Buy one reading, if you want it
The paid reading costs credits, shows its price before you agree to it, and reports what it refused alongside what it found.

Analyze your manuscript

What is manuscript analysis?

Manuscript analysis is the pass you make over a finished draft before anyone else sees it, looking for the things that are wrong in a way you can point at. Not whether the story works, which is a judgement, but whether a chapter is empty, whether a name is spelled two ways, whether the book gives two different figures for the same thing, and what the shape of the prose actually is once somebody counts it. Most of that is arithmetic, and arithmetic is checkable: if a report says a sentence runs to sixty words, you can count the words. BookFoundry splits the capability along exactly that line. Everything that can be proven by counting is free and runs instantly, because no model runs and there is nothing to charge for. The one part that needs reading rather than counting is priced on its own, and it is deliberately narrow.

What the free layer reports

Four things. The statistics of the book: word and chapter counts, how evenly the chapters run, average sentence length and how much it varies, how much of the text is dialogue, the words that repeat, and the filler words that carry no weight. The publish readiness check: an empty chapter, two chapters holding the same text, a contents or copyright page that arrived as a chapter during an import, raw formatting characters showing as text, an untitled chapter, two chapters claiming the same position. An estimate of where the free sample a shopper reads probably ends, with the sentence sitting at that boundary. And candidates: places where the book gives two different numbers for what looks like the same thing, or writes one name more than one way. All of it is free, in every language BookFoundry writes books in, and a measure with no word list for your language returns nothing and says so rather than guessing.

Why a candidate is not an error

Two different years attached to the same subject may be two different events. Sara and Sarah may be two people. A tool that calls either one a mistake is wrong often enough that authors stop reading its output, and once that happens the real findings go unread too. So the number and name results are presented as candidates: both passages are quoted, in your own words, and you decide. The same principle runs the other way. When the analysis finds nothing, that is stated as the narrow thing it is. It means nothing countable went wrong. It is not a verdict on the writing, and it is never described as a clean bill of health.

Where your free sample probably ends

Stores let a shopper read the opening of a book before they buy it. How much they show is not published, differs between stores and between books, and changes without notice, so the position Analysis marks is an estimate and is labelled one everywhere it appears. We make no claim to know any store's number. The value does not depend on the percentage being exact: it comes from seeing, probably for the first time, the actual sentence a shopper stops at, and the text that sits just past it, which is what they would be buying the book to read next.

The one paid reading, and the two things it looks for

One reading costs credits, and it detects exactly two things. A section that leaves the reader holding nothing: the passage states things but does not arrive anywhere. And a section that covers ground the book already covered, in much the same terms. Both are quoted so you can read the passage and disagree. That is the whole list. It is short because a reading that claims to find everything finds nothing reliably, and because each of these two was worth shipping only once it could quote the evidence for itself. The price is on screen before there is a button to press, starting the reading takes two deliberate actions, and what the reading refused is shown next to what it reported, because a dismissal you cannot see is indistinguishable from a lucky guess.

What it deliberately does not do

It does not rewrite your book. Not a sentence, not a word, on any screen, free or paid. It does not score your writing, rank it, or tell you whether a chapter works: those are judgements, and a judgement dressed up as a statistic is worse than no statistic. It does not predict sales or how readers will react, and it makes no claim about what any store will show. It does not publish an accuracy percentage, because the only figures ever measured for the paid reading come from a synthetic test set, and a number like that on a page like this would be read as a promise about your book. And where a measure cannot speak about your language, it returns nothing and tells you it could not answer, rather than producing a confident answer from a table it does not have.

Frequently asked questions

The free layer reports four things: the statistics of the book, publish readiness such as empty or duplicated chapters and raw markup showing as text, an estimate of where a shopper's free sample ends, and candidates where two numbers or two spellings of a name disagree. One paid reading looks for exactly two more things.

The whole deterministic layer is free for the whole book, because it is arithmetic and no model runs. One reading costs credits, and its price is shown before you agree to it. You can also try a bounded version on a single pasted page with no account at all.

No. It reports and stops. Every change to your book stays your decision, and no screen in Manuscript Analysis writes into your text.

It does not know, and it says so. Stores do not publish how much of a book they show and it differs by store and by book, so the boundary is an estimate. What is useful is not the percentage, it is seeing the sentence that sits there.

It means nothing countable is wrong: no empty chapters, no duplicates, no scaffolding imported as a chapter. It is not a judgement on the writing and it is not a clean bill of health. Reporting it as one would be the easiest lie this product could tell.

Yes, in all 18 languages BookFoundry writes books in, including Arabic with full right-to-left support. Where a measure needs a word list for a language and there is not one, it returns nothing and tells you rather than reporting the language's own grammar as your overused words.

AI book editorThe AI book editor that reads your whole bookFive sections, seven readings that call no AI, and paid readings priced before they run. Everything the editor does with a finished book, on one page.

Ready when you are

No setup, no learning curve. Start in your browser and hold a finished book today.