Most authors are not asking whether to use AI at all any more; they are asking how to use it sensibly. That is a better question, because it depends on where in a project AI sits, and what it is being asked to do.
This guide sets out a practical framework: the uses that tend to genuinely help, the uses that tend to cause problems, and the current platform and disclosure context worth knowing before you settle on your own approach.
What are the genuinely good uses of AI for authors?
- Planning: generating questions and options for premise, structure and character, covered in can AI help me plan my novel?
- Research preparation: generating questions to investigate, never facts to trust directly
- Organisation: turning scattered notes into an ordered outline or to-do list
- Manuscript analysis: flagging repetition, pacing issues or inconsistent terminology
- Continuity: helping build and query a series bible, covered in can AI check continuity in a book?
- Editing support: a first pass on grammar, spelling and consistency alongside human editing
- Marketing: brainstorming content ideas, taglines or description drafts to refine yourself
- Metadata support: generating draft category or keyword ideas to check against platform requirements
What are the uses worth avoiding or treating with caution?
What does this mean?
Blind fact reliance
Treating a model’s confident-sounding answer as a verified fact, date, statistic or citation without independent checking.
What does this mean?
Wholesale rewriting
Replacing large sections of your own drafted prose with generated text, which risks flattening voice and introducing errors you did not write and may not catch.
What does this mean?
Voice replacement
Letting generated prose stand in for your own distinctive style across a manuscript, rather than using AI to question and refine what you have written.
What does this mean?
Uploading sensitive material without understanding terms
Pasting unpublished manuscripts, personal material or anything covered by a confidentiality agreement into a tool without checking how that service stores and uses input.
What does this mean?
Invented citations
Accepting a generated reference, quotation or source without confirming it actually exists, since models can produce plausible-sounding citations that are entirely fabricated.
What is a simple framework for deciding how to use AI?
What do current platform obligations look like?
At the time of writing, Amazon KDP asks publishers to disclose AI-generated content when publishing or updating a title, while not currently requiring disclosure of AI-assisted content that you wrote and then refined with AI tools 1. Other platforms and distributors may define and require disclosure differently, and these policies continue to change, so check the current wording for any specific platform before you publish. This is general information for authors, not legal advice. Copyright and platform rules around AI are changing quickly and vary by country; if a specific project depends on getting this right, check current official guidance or ask a qualified professional.
Should I set my own AI policy for a project?
Deciding, before you start, which uses you are comfortable with, planning support but not generated prose, for instance, tends to make individual decisions easier and more consistent as a project develops. It also makes it simpler to answer honestly if an agent, publisher, editor or reader ever asks how a book was produced.
Questions to answer when setting your own AI policy
- Which stages of this project (planning, drafting, editing, marketing) will I use AI for?
- Will I ever use generated prose directly in the manuscript, or only for planning and questions?
- How will I verify anything factual that AI produces?
- Have I checked the data and privacy terms of any tool I plan to use with unpublished material?
- Do I know the current AI disclosure requirements for every platform I plan to publish on?
- Am I comfortable disclosing how AI was used if asked by an editor, agent or publisher?
Common mistakes
- Using AI without any personal policy or boundaries. Deciding case by case, under deadline pressure, tends to produce inconsistent and sometimes risky choices.
- Treating "AI-assisted" and "AI-generated" as the same thing. Platforms increasingly distinguish the two for disclosure purposes; know which applies to your manuscript.
- Assuming a confident answer is a verified one. Facts, quotations and citations from AI need independent checking every time.
- Letting AI use expand without noticing the voice has changed. Periodic read-throughs help catch drift towards generic, AI-flavoured prose before it spreads through a manuscript.
Frequently asked questions
Is there one correct way for authors to use AI?
No single approach fits every author or project. The framework here, using it to think, check, organise and question rather than decide, verify itself or become the author, applies broadly, but the specific boundaries you set are a personal decision.
Do I need to disclose AI use even if a platform does not require it?
You are not obligated to beyond a platform’s actual rules, but being straightforward if an editor, agent or reader asks directly is generally the safer position.
Does using AI at all reduce a book’s value or authenticity?
Not inherently. Planning and organisational uses are widely accepted; the more a book’s final prose originates from generated text rather than the author, the more real questions arise about voice, originality and disclosure.