Prompts for Books: Plan, Draft, and Revise With AI

Good AI prompts for writing a book anchor the model to your specific story, characters, and intentions rather than a generic version of what a book like yours might contain. The most effective prompts supply the project context before making a request, so the AI is working with your material instead of inventing its own.

The prompts below are templates you adapt to your project. Tools like fenbs keep your book's project notes, writing decisions, and session progress in one place so each new session starts from where the last one ended rather than from scratch.

Write a project note before you start prompting

Before using any of these prompts, write a short project note for your book. Every useful prompt will reference it. This note does not need to be long — a paragraph on what the book is, who reads it, and what it should achieve is enough to anchor a session.

My book is [working title]. It is a [genre/type] aimed at [audience]. The central subject or story is [one sentence]. The tone is [tone]. The main characters or voices are [names and brief descriptions]. Key things it must include or avoid: [constraints].

Paste this note at the start of any session. When you revise it as the book develops, save the new version so future sessions use current information. A stale project note is worse than none — it sends the AI in the direction you were going six months ago, not where you are now.

Prompts for planning and outlining

Planning prompts work best when they push back as well as generate. Ask the AI to surface problems with your structure rather than just fill it out. An outline that has never been stress-tested is harder to write than one that has already survived a few pointed questions.

Using the project context above, suggest a chapter-by-chapter outline for this book. For each chapter, give the main event or argument, what the reader should understand by the end, and any unresolved question it should leave. Then tell me where the structure is weakest and why.

Look at this outline: [paste outline]. Identify the three places where a reader is most likely to lose momentum or interest, and suggest what would fix each one. Do not rewrite the outline — just flag the problems.

Keep your finalized outline in your project notes. When you revise a chapter's purpose, update the outline there so later drafting prompts stay consistent with what the book is actually becoming. An AI that drafts Chapter 7 without knowing what changed in Chapter 4 will write something that does not fit.

Prompts for drafting chapters

Drafting prompts need to constrain the AI to your story's established facts. Without constraints, a model will invent details that contradict what you have already written. The more specific your constraints, the less work you spend correcting the draft later.

Write a draft of Chapter [N]: [title]. The chapter should: [main events or arguments]. It must be consistent with these established facts: [key facts]. The tone should match [tone description]. End with [intended ending state — what the reader knows or feels]. Length: approximately [word count] words.

Draft the scene where [description]. The characters present are [names]. What we know so far about each of them: [brief]. The setting is [setting]. The scene needs to establish [what it establishes]. Do not introduce new characters or contradict [specific constraint].

After each draft, note what works and what does not before moving to the next one. A short note — "the opening was too slow; cut to the dialogue sooner" — takes thirty seconds and prevents you from repeating the same structural choice in the next chapter. These notes also help you develop a drafting style with this particular AI that gets more reliable over time.

Prompts for checking continuity

Continuity errors are one of the most common AI writing problems. The model does not remember what it wrote in earlier sessions unless you tell it. A continuity check prompt is most useful after you have a working draft of several chapters, and it works best when you supply the summary of established facts yourself rather than asking the model to recall them.

Here is a summary of what has been established in Chapters 1 through [N]: [paste summary]. Now read the draft of Chapter [N+1]: [paste chapter]. Identify every place this chapter contradicts, ignores, or is inconsistent with what is already established. List each inconsistency separately.

Here is what we know about [character name]: [description, voice, history, stated motivations]. Read the following passage: [paste passage]. Identify any dialogue, action, or statement that feels inconsistent with this character as described.

Keep a running list of established facts as you draft — character names and descriptions, dates, locations, decisions people make. This list becomes the input for continuity checks. Without it, you are asking the model to check a chapter against a memory it does not have.

Prompts for revision

Revision prompts work best when they are specific about what to improve. A prompt like "make this better" produces generic feedback. "Identify the three sentences that are hardest to follow on a first read and suggest clearer versions" produces actionable changes you can evaluate and accept or reject.

Read the following passage: [paste]. Identify the three sentences that are hardest to follow on a first read. For each one, explain why it is confusing and suggest a clearer version that preserves the meaning.

Read this chapter: [paste]. Identify the two sections where the pace slows the most. For each one, say whether the slowdown is a problem or whether it serves a purpose (for example, a deliberate moment of reflection). If it is a problem, suggest what to cut or compress.

When a revision note becomes a rule — "always use short sentences for action sequences," "never head-hop in this book" — record it as a decision so it applies to every future session rather than having to be re-established each time. A decision recorded once travels with the project; a decision kept only in your memory gets forgotten after three sessions.

When AI suggestions do not fit your vision

AI will sometimes generate something technically correct but wrong for your book. A suggested chapter structure that works for a different kind of story, dialogue that is grammatically fine but does not sound like your character, or a revision that clarifies the sentence but loses the voice. This is expected, and it does not mean the prompt failed.

The useful response is to be specific about what went wrong. "That structure works for a thriller but my book is literary fiction — try again with longer, slower scenes" gives the model something to work with. "This does not feel right" does not. Over a few sessions you learn which types of prompt produce useful drafts and which need more constraint, and you update your templates accordingly.

How to do this with fenbs

1. Create a free account. Set up a project for your book.
2. In the project, add your book project note as a context note. Add a task for each chapter, with a note on its purpose and its current state.
3. In Skills and prompts, save your most useful prompt templates by stage: outline, draft, continuity, revision.
4. Connect an AI app from Integrations and ask it to read your project context and the relevant prompt skill at the start of each session.
5. Before you end a session, update the task for the chapter you worked on and save progress: what you drafted, what you decided, and what comes next.

fenbs does not write or store your manuscript. It keeps your project notes, decisions, and session notes so you always have a current starting point. Your drafts live in your writing app; fenbs keeps the information that makes the next session productive rather than a reset.