AI and Data

AI features are disabled when your organization is created. They stay disabled until an admin deliberately enables them.

This is enforced on our servers, not by hiding a button. While AI is off, no AI action will run — including through our API, and including for an admin. Nothing about your book is sent to a model provider.

An admin can turn it off again at any time. Doing so doesn’t affect anything you’ve already created.

There is no background job that reads your work. Nothing runs on save, nothing runs on a schedule, and nothing indexes your manuscripts for any purpose.

Every AI action in BookWorkz is a button you press. If you never press one, your manuscript never reaches a model.

Move In, where you bring in your backlist, files and sales reports, doesn’t use AI. It sorts what you bring in by rules we wrote, and nothing is sent to an AI provider during Move In. Whether AI features are used elsewhere is up to you: they stay off unless an admin turns them on.

We’d rather tell you which actions read your prose than claim none of them do.

What is sent depends on the action.

Planning and describing actions send your story materials. Generating chapter beats, jacket copy, character bios, settings, lore, social posts (including a YouTube video’s title, description and tags) and a first chapter draft sends the outline and notes they build on — character notes, settings and lore — and the brief you type. Drafting, continuing and rewriting a chapter or a passage, and generating chapter beats, also send the style guide that applies to your book. Generating an outline sends the brief you type and your pen name’s voice profile. Plan feedback sends your plan. Building a timeline sends your plan’s chapter summaries, your chapter titles, and the names of your characters and places — not their notes.

Editing and rewriting actions also send the prose they work on. A developmental edit and a chapter rewrite send that chapter. A passage rewrite sends the passage and some of the text around it. Rewriting several chapters together sends the text of each chapter you selected, in book order, with each chapter’s title to mark where it starts, and some of the text just before and after them. A copy edit and a line edit also send the style guide that applies to your book.

Your style guide. When you draft, rewrite, edit or write copy, we send the style guide that applies to that book or pen name: the house rules your publisher, pen name and series set, and the book’s own. You see and change them under House standards.

Address with AI, on a margin note, sends the passage the note is on, the paragraph before and after it, that one note’s text, any guidance you add, and your book’s voice settings, and the style guide that applies to your book. It never sends your other notes. It goes through OpenRouter to OpenAI by default; if the model declines and you choose to try another, it goes to Anthropic, then to Together.

An editorial pass sends each chapter it reads, any guidance or instructions you add, and your book’s voice settings, and the style guide that applies to your book; a continuity pass also sends your characters, settings, lore and timeline. It never sends your notes. Its comments come back as notes in your margin, private to you. It goes through OpenRouter to OpenAI by default; if the model declines and you choose to try another, it goes to Anthropic, then to Together.

Suggest a title, on a chapter, sends the beginning and end of that chapter (or its planned beats, before it has any text), its current title and subtitle, the titles of the chapters either side of it, your book’s voice settings, and the style guide that applies to your book. It never sends your notes. It goes through OpenRouter to Anthropic; if the model declines and you choose to try another, it goes to Together.

Actions that summarize or learn from the whole book send the whole manuscript. These are a synopsis, an outline built from your chapters, and the character and setting notes built from your chapters. An editorial pass sends each chapter you choose, one at a time, so a pass over every chapter sends the whole manuscript.

A voice profile sends the pen name’s description and any sample writing you paste in.

Image actions send the image: photos of handwritten outline pages, photos of a style guide’s pages, and newsletter or social images when you ask for alt text or a caption.

Reading a style guide into a brand guide sends the style guide you choose: a PDF as it is, a Word document’s text, or photos of its pages. It sends nothing else from your guides or your books, and it goes directly to Anthropic. What it reads comes back as a draft for you to check, section by section; nothing goes into the guide until you choose. Style guides you attach or read are stored privately with the brand guide, where only people who can see that guide can open them. Photos you read are kept together as one PDF. A file is deleted when you replace or remove it.

Reading a character bio with AI sends the bio you choose: a PDF as it is, a Word document’s text, or photos of its pages. It sends nothing else from your book, and it goes directly to Anthropic. What it reads comes back as a draft for you to check, field by field; nothing is saved until you choose. Files you attach to a character are stored privately with that character, where only people who can see the character can open them, and are deleted when you replace or remove them, or when the character is deleted for good.

Reference photos you add to a character are stored privately for your team and are never sent to an AI model. No AI action reads them, and they are kept apart from the character’s notes, so an action that sends the notes doesn’t send the photos.

Idea generation sends the prompt you type, your pen name’s voice profile, and the titles of up to 100 of your existing ideas, so it doesn’t suggest one you already have. It sends no manuscript text.

Your voice settings are sent as you wrote them: your pen name’s voice profile, your series’ and books’ voice notes, and your pen name’s Marketing voice, each with the actions described under “It’s built for writers.” A pen name’s sample prose is used only to draft its voice profile and is never sent with anything else.

In the writing editor, a line beginning ‘Using:’ shows what will be sent before you are charged.

Publishing a social post sends nothing to an AI provider. The post goes to the platform you aimed it at, as you wrote or approved it.

AI output arrives for review. It is never written into your manuscript automatically.

A chapter, an outline, a character bio, a synopsis — you see it first, and you accept it, change it, or throw it away. The version that lands in your book is the version you approved.

BookWorkz has no models of its own, and your manuscripts are never used to build one.

We use third-party model providers to run AI actions. Every provider we use is named in Section 6 of our Privacy Policy, and where we reach a provider through a routing service, both the routing service and the provider that performs the processing are named there. We send AI requests only to providers that are barred from using your content to train or improve any model, either by a restriction we set on the request or by the provider’s own commercial terms.

We do not authorize any provider to use your work for training on our behalf. If we add a provider, or change how one is reached, the Privacy Policy is updated before anything is sent to them.

We claim no ownership of AI output and take no license in it. As between you and us, it’s yours, the same as anything else you write in BookWorkz.

One thing worth knowing, because it’s law rather than our policy: in several jurisdictions, material generated by AI without meaningful human authorship may not be eligible for copyright protection. What that means for your book depends on how you use these features and on the rules of the retailers and distributors you work with. You’re in the best position to judge that, and to make any disclosures your contracts require.

AI usage is metered in credits, billed separately from your subscription. The credit cost of every action is shown before you run it — you never spend without seeing the number first.

Admins can set a monthly spending ceiling. Automatic top-up is off unless someone turns it on. If your balance runs out, AI actions pause and the rest of BookWorkz keeps working.

Export is never metered. Producing a DOCX, an EPUB, or a print-ready PDF costs no credits on any plan, ever. Formatting your book is not an AI feature and we don’t charge for it as one.

When you draft, continue or rewrite prose, generate chapter beats, an outline, character bios, settings or lore, or suggest a chapter title, the AI receives your narrative voice, in layers: your pen name’s voice profile (point of view, tone, style notes and comparable titles), then your series’ voice, then the book’s own. Each layer adds to the one before it and can change its point of view, tense and tone. You can see the exact text the AI receives on the book’s Book Bible, under “What the AI will be told.” A character, setting or lore item shared across a series receives the series’ voice. A character’s own voice notes are sent as that character’s dialogue and inner voice, never as the book’s narration. Generating ideas uses your pen name’s voice profile alone, and a synopsis receives only the book’s tone and genre. Jacket copy, social posts and newsletter suggestions receive your pen name’s Marketing voice instead. Actions that review your work rather than write in your voice — outline analysis, the developmental, line and copy edits, Get feedback, building a timeline, reading photographed pages, and reading a style guide — don’t receive it, though the content edit is told the book’s point of view and tense so it can check them. Either way, the prompts are written for fiction rather than for general-purpose text. A cozy mystery and epic fantasy are not the same job, and the prompts know it.

The output is meant to sound like your book, not like the average of all books. It won’t always. That’s why nothing saves without your approval.

Model providers set their own acceptable use policies, and what a given model will and won’t write varies between them. That’s their decision, not ours.

Our own rules are in Section 7 of the Terms of Service and they are narrow: content that is unlawful, that infringes someone’s rights, or that depicts the sexual exploitation of minors. We write for genres that go to dark places, and we’re not in the business of policing what your book is about beyond that line.

Where we offer a choice of model for an action, the choice is yours.

Running AI has a real energy cost. We give 1% of our revenue to offsetting the environmental cost of the AI we run — revenue, not profit, so it starts the day someone subscribes rather than the day we’re comfortable.

AI output can be wrong, incomplete, or unsuitable. It isn’t editorial, legal, or professional advice, and it isn’t a substitute for an editor. Read it before you rely on it.