What authors ask before signing up.

No. Your manuscripts never enter our training data — this is a hard contractual commitment with every model vendor we route through, not a marketing claim. We log token counts for billing; we never keep your prose in any AI pipeline.

Full studio access: write, upload existing manuscripts (DOCX/TXT/MD/RTF), use Grove for chat, chapter reviews, entity extraction, and Author Directives. Heavy whole-book actions (full-manuscript story doctor, blueprint review on giant uploads) unlock with the paid plan. Export via Publishing Studio or the archive ZIP unlocks once the first paid month begins. Card up front so the transition to paid is seamless — cancel before day 7 and you're not charged.

Cancel before day 7 and your card is never charged. Your books remain encrypted in your account; sign back in any time. One important honest note: if you've already exported your work via Publishing Studio or the archive ZIP, you've taken the value the subscription was for, and the 30-day money-back guarantee no longer applies after export. This keeps trial pricing sustainable for actual authors using the tool.

It covers subscriptions where you haven't already exported your manuscripts. Once you use Publishing Studio or the export archive, the subscription's value has been delivered and the guarantee no longer applies.

You keep everything. Your encrypted library remains in your account — sign back in any time to read or export. While you're subscribed, you can download your entire library as an archive whenever you want. We don't hold your work hostage — but we do gate the first export to subscribers who've completed at least one paid month, so trial abuse doesn't subsidize itself on the backs of paying authors.

Different center of gravity. Those tools focus on generating prose — autocomplete, expand, full-draft generation. Scribegrove is built around an editor: Grove reads your whole manuscript, flags pacing, continuity, and voice drift with chapter-level anchors, and never changes a word without your approval. Then it covers the parts of book-writing the generators skip: a built-in beta-reader loop (share a private reading link; notes flow back into your workspace), a real phone reading experience with AI margin notes, and validated EPUB 3.3 + ONIX 3.0 publishing exports.

Yes. Ask Grove to hand you one of your characters ("let me talk to Marlow," "interview my villain") and it steps into that character and answers in the first person — grounded in your canon, so they only know what they'd plausibly know at that point in the story (no plot spoilers, no other-POV secrets), and honoring your book's content rating. It's a discovery space: it never edits your prose. The one thing it does remember — when you correct how a character spoke ("too formal," "she's blunter"), Grove appends that as a rule to the character's voice notes behind an Apply button, so every future scene the book generates stays in character. Your correction becomes canon, transparently and only with your approval.

Grove is tuned for fiction with persistent memory of your series world (characters, locations, magic systems, rules), your Author Directives ("more horror," "first-person POV," "keep dialogue snappy"), and your book's Spice rating. ChatGPT and Claude are general-purpose chatbots with no per-book memory between sessions — every conversation starts cold. We use the same class of upstream models for heavy work, but the wrapping is what makes a novel-writing tool feel like a novel-writing tool.

Grove doesn't just answer one question at a time. It autonomously plans + runs multi-step reviews against your work. The Story Doctor reads your whole manuscript, identifies pacing / continuity / voice / structural issues, and emits an inline report with chapter-level anchors so you can preview each finding before applying anything. Blueprint Review (on manuscript imports) chunks the document, extracts characters / locations / systems / canon items, and synthesizes a series-ready entity graph. Insights continuously surfaces drift (name spelling, timeline contradictions, voice shifts) without being asked. Every output is preview-then-apply — never an automatic rewrite.

Yes. Scribegrove is AI-assisted — you write and shape the book; Grove edits, checks continuity, and suggests, but never publishes for you or claims your work. U.S. copyright rests on human authorship, so a book you substantially author is yours. We never take a cut of your sales, and we never label your work.

Yes. DOCX, TXT, Markdown, and RTF imports are supported in the new-book wizard. Scrivener users should compile to DOCX first (Scrivener → Compile → Microsoft Word .docx). The wizard previews the parsed structure (chapters, scenes, headings) before commit so you can adjust before anything is final.

Yes — Scribegrove runs in any modern browser, so macOS, Windows, Linux, ChromeOS, iPad, iPhone, and Android all work today with nothing to install. The phone experience is built for reading and noting: a paged reader, select-to-annotate AI actions, and voice notes. A desktop companion app is on the roadmap — same app as the web, with offline writing and note-taking (AI features will still need a connection).

No. Scribegrove is a fully managed product — we handle model routing, capacity, and updates so you never have to think about API keys or per-call billing. Your subscription covers it; that's the whole point.

Yes. Set your book's Spice rating (1-5) and Grove honors it strictly — at the rating you set, without escalating and without sanitizing. Explicit content routes only through model paths that support it contractually, so the rating is product policy rather than a coin flip against a general-purpose content filter.

Every export from Publishing Studio ships with ONIX 3.0 metadata + Schema.org accessibility features by default. Compliant with the June 2025 EU Accessibility Act and Amazon's ONIX 2.1 sunset (March 2026). No extra checkbox to remember.

Your manuscripts are encrypted at rest in our cloud with per-user keys (AES-256). Your account is recoverable the way you'd expect — standard email-based password reset, nothing to lose — the same trust model as Google Docs or Notion. We keep billing-related logs (which AI calls ran when, how many tokens), and you can export your full library any time while subscribed.