How to Use AI to Help Write a Book: Outlines, Drafts, and Continuity
ChatGPT and Claude can help you plan, draft, and revise a book chapter by chapter. Here is how to get useful work from them while keeping your voice and your editorial decisions.
Updated 6 min read
AI tools like ChatGPT and Claude can help you write a book by generating outlines, drafting scenes, suggesting structure, and checking continuity. They work best as a collaborator: you supply the ideas, decisions, and voice, and the AI handles the first heavy pass. Tools like fenbs keep your manuscript brief, chapter progress, and editorial decisions in one place so each new session picks up where the last one stopped.
The useful workflow is not "ask the AI to write the book." It is a loop: give the AI the current chapter plan, let it draft, revise, record the decisions you made, then brief the next session from that record. That loop is the same whether you use ChatGPT, Claude, or both alternately.
What AI can and cannot do for your book
AI assistants are good at generating first drafts, proposing chapter structures, writing description and dialogue, and suggesting revisions. They can move quickly through an expository section, help you work around a plot problem, and read an existing chapter for internal consistency.
They are less reliable on facts, on long-term continuity across sessions, and on your specific voice. A ChatGPT draft of chapter twelve may contradict something you established in chapter three, because the session does not hold those earlier chapters in its context. Always verify factual claims against a primary source, especially in nonfiction.
For novels about specific historical events or nonfiction that cites research, treat AI as a drafting tool, not a research tool. Draft with it, then fact-check what it produced.
Build a manuscript brief before your first session
A manuscript brief is a document — or a set of notes — that tells any AI session the essentials of the book. It is the most useful thing you can prepare before starting, and the most important thing to update as the book evolves.
- The book's premise: one or two sentences on what it is and who it is for.
- Characters or key figures, with a short note on how each sounds and what they want.
- The timeline or structure: how the book is divided and what each section covers.
- Decisions already made: point of view, tense, what the reader learns and when.
- Rules for the world, tone, or style that the AI must not break.
Paste the manuscript brief at the start of each new session. It prevents the AI from inventing details that contradict earlier choices. Record any change to the manuscript brief as a decision, not just an edit, so you can trace why the book evolved in a particular direction.
Create your outline with AI
Start with the manuscript brief, then ask the AI to propose a chapter breakdown. A good prompt gives the premise and the intended ending and asks for chapter summaries with one main action and one turn in each. Review the outline before writing: move sections, combine chapters, and cut anything that repeats.
- Share the manuscript brief with the AI and ask for a chapter-by-chapter outline.
- Ask it to note structural problems: chapters that do the same job, gaps in the sequence, or a resolution that arrives too early.
- Work through the feedback and record your decisions — which suggestions you accepted, which you changed, and why.
- Save the approved outline. Every future session should start from this, not from scratch.
Keep the outline as a decision, not a draft. When you change a chapter plan, note what changed and why. The AI brain guide explains the broader idea of keeping useful decisions outside any one chat session.
Maintain continuity across chapters
Continuity is the hardest part of AI-assisted book writing. A session that covers chapter twelve does not automatically know what you established in chapter three. A character's detail, a plot rule, or a voice choice can shift without anyone noticing until a reader flags it.
Keep a running continuity sheet alongside your outline: the characters who appear in the current chapter, the facts already established, and any rules the scene must follow. Before each drafting session, paste the relevant section. After drafting, ask the AI to check the result against those facts before you move on.
Record rule changes as decisions. "The protagonist does not speak in the final chapter" is a rule. "The timeline moved back by one year" is a decision. Both change how later chapters must be written, so both need to be somewhere the next session can find them.
Keep your manuscript brief and chapter progress across every session
fenbs keeps your manuscript brief, chapter progress, continuity decisions, and editorial rules in one place. Connect Claude or ChatGPT over MCP at /integrations and each session reads the current state of the book before it writes. Before you stop, save where the chapter stands and the decisions you made, so the next session does not start from zero.
Create a free accountFree for one person · No credit card · Works with Claude, ChatGPT, Cursor and more
Write first drafts with AI support
For a chapter draft, paste the manuscript brief, the outline entry for that chapter, and the relevant continuity notes. Then ask for a draft. The first pass is usually too generic — that is expected. Your job is to revise, not to accept it wholesale.
Good uses for AI drafts include action sequences, transitions between scenes, expository paragraphs, and any section you are stuck on. Let the AI write a rough version, take what is useful, and rewrite in your voice. If the AI keeps defaulting to a style you did not ask for, add a voice note to your brief.
Revise and refine with AI
AI is effective at revision when you give it a specific task. "Check this chapter for pace: flag any paragraph that slows the scene without adding information" is more useful than "revise this chapter." Give it one revision task at a time and review each pass before asking for another.
For voice, revision often means undoing what the AI wrote in the first pass. A chapter that reads like a generic AI response will feel different from the rest, and readers will notice. Use the draft as a scaffold, then rebuild the sentences in your own rhythm. Guide the next session's work with the how to write a task for an AI agent approach: precise scope, clear output format.
How to do this with fenbs
Set up a fenbs project for your book, treat each chapter as a task, and record your key editorial choices as Decisions and rules. Connect your AI app over MCP so it reads the current brief before drafting.
- Create a free account and start a personal kanban board for your book project.
- Add the book as a project. Write a short description with the premise, structure, and key decisions so far.
- Add each chapter as a task. Note its outline entry, any continuity facts that apply, and its current status.
- Record your editorial decisions — point of view, voice rules, approved structural changes — in Decisions and rules so every AI session reads them first.
- Connect your AI app from Integrations. At the start of each session, ask it to read the project context and the current chapter task. Before you stop, ask it to save progress and note what still needs writing or revision.
fenbs does not host your manuscript or replace a writing tool. Keep the draft in your writing app. Keep the decisions, progress, and brief in fenbs so any connected AI session starts with the current plan rather than asking you to re-explain everything.