How to Document AI Use Before You Publish a Nonfiction Book on KDP

For a practical nonfiction book, document AI use as you draft, not after the upload screen asks about it. Keep a simple record showing what was human-written, what was AI-assisted, what was AI-generated, what you verified, and which final files contain that material. On KDP, confirm the current AI disclosure requirements during title setup, because Amazon’s settings and definitions can change.

Start with the distinction KDP cares about

The most important documentation decision is whether your use of AI is AI-assisted or AI-generated. In KDP’s current guidance, AI-assisted content generally means you created the content yourself and used AI tools to edit, refine, brainstorm, check grammar, or improve workflow. AI-generated content means an AI tool created text, images, or a translation, even if you later edited that output.

For a practical nonfiction book, this distinction can appear in several places. A chapter outline suggested by AI may be AI-assisted if you wrote the manuscript yourself. A paragraph, case study, chart explanation, glossary entry, recipe, exercise, or cover image produced by AI may be AI-generated if the tool created the expression you used in the final book.

Do not rely on memory at the end of the project. If you are using AI while drafting, build a record that follows the manuscript. The goal is not to create a courtroom-sized archive. The goal is to have enough detail to answer KDP’s disclosure questions truthfully and to correct your files later if you discover a problem.

Create an AI use log before the manuscript becomes complicated

A practical AI use log can be a spreadsheet, document, or project note. Include the date, tool name, book section, type of use, and your decision about the output. Useful labels include brainstorming, outline support, grammar edit, rewrite suggestion, fact-check prompt, generated text, generated image, generated translation, and rejected output.

Add a column for the final file affected. This is the part many authors miss. KDP decisions are tied to the files and metadata you upload: manuscript, cover, interior images, translated edition, book description, subtitle, keywords, categories, and sometimes A+ Content. If AI output appears only in discarded drafts, note that it was not used in the final book.

For nonfiction, add a verification column. If an AI tool suggested a fact, example, citation, legal point, medical statement, historical detail, or technical step, record how you verified it or whether you removed it. KDP compliance is not only about the AI disclosure field. You are still responsible for the accuracy, legality, and reader experience of the published book.

Keep the log in plain language. You do not need to save every prompt unless doing so helps you track rights, sourcing, or quality. A short entry such as “Chapter 4 checklist, AI generated first draft, rewritten and verified against author process notes, appears in final manuscript” is more useful than a folder full of unlabeled screenshots.

Map AI decisions to the files you will upload

KDP title setup asks for information about the book you are publishing, so your documentation should follow the finished files. For the manuscript, mark any AI-generated sections that remain in the final interior. For a practical nonfiction book, these may include summaries, worksheets, quiz questions, example scenarios, definitions, or chapter introductions.

For the cover, record whether any image, illustration, background texture, icon, or composite element was AI-generated. If a designer used AI as part of the process, ask for a written summary before you approve the final cover. You need to know enough to answer KDP’s AI content questions and to understand what rights the designer is giving you.

For interior visuals, keep an asset register. List each chart, diagram, photo, illustration, table, worksheet, or icon. Note whether it was human-created, licensed stock, public domain, commissioned, AI-assisted, or AI-generated. Also keep the license or permission record where applicable. This matters because later editions, large-print versions, workbooks, and translations often reuse the same assets.

For translations, record whether AI created the translated expression that remains in the final edition. Human review or extensive editing of a generated translation does not change that origin. A human-created translation refined with AI is a different case; ask the translator to describe the process and identify any generated passages.

Separate source documentation from AI documentation

A nonfiction book needs source control whether or not AI is involved. AI tools can make this more important because they may produce confident statements that still need independent verification. Keep a source log for factual claims, instructions, professional advice, quotations, statistics, studies, and examples drawn from real people or organizations.

Check the original source behind any factual claim suggested by AI. If support cannot be found, remove the claim, narrow it to what the evidence actually establishes, or obtain appropriate review. Relabeling an unsupported factual assertion as personal opinion does not make it reliable, especially in health, law, finance, or safety guidance.

Your source log and AI use log serve different purposes. The source log explains why the content is accurate or supportable. The AI use log explains how the content was created. Together, they help you answer reader concerns, correct errors, update future editions, and keep your KDP disclosure consistent with the actual book.

Plan front matter and author notes without overpromising

KDP’s AI disclosure is handled in the publishing workflow, but you may also choose to mention AI use inside the book. This is not always required, and you should confirm current KDP policy before treating any interior statement as sufficient. An author note can still be useful for transparency, especially in practical nonfiction where readers rely on your judgment.

Keep any disclosure inside the book specific and modest. For example, you might say that AI tools were used for editorial assistance, outline testing, or draft review, while the author remains responsible for the final content. If the book contains AI-generated exercises, illustrations, or summaries, identify the type of material without turning the note into marketing copy.

Avoid statements you cannot support, such as claiming that AI made the book more accurate, more researched, or guaranteed to be original. If you used professional reviewers, editors, designers, or subject-matter readers, describe their role only if it is true and you have permission where needed. Transparency should reduce confusion, not create new compliance risk.

Use your records when completing KDP metadata and disclosures

When you reach KDP title setup, use your AI use log to answer the AI content questions based on the final manuscript, cover, and other files you are uploading. Amazon’s Help pages on AI content, content guidelines, and metadata guidelines are the first places to check because the platform can update fields, definitions, and enforcement practices.

Metadata also needs restraint. Do not use the description, subtitle, keywords, or categories to make unsupported claims about AI, expertise, bestseller status, endorsements, or results. Do not stuff keywords such as “AI,” “KDP,” or tool names into metadata unless they accurately describe the reader-facing content of the book.

Record what AI created, what it helped revise, and where that material appears in the book. This gives the disclosure review a concrete starting point and reduces the need to reconstruct the workflow immediately before upload.

After publishing, keep the records with your book files. If you upload a revised manuscript, replace a cover, add illustrations, or issue a translated edition, revisit the AI disclosure. A change in the final files can change the answer you gave during setup.

A simple documentation workflow for a practical nonfiction book

Before drafting, create three files: an AI use log, a source log, and an asset register. Name them with the book title and date so they do not get separated from the manuscript. Decide in advance how you will label AI-assisted work versus AI-generated work.

During drafting, update the AI use log whenever AI output becomes part of the manuscript or assets. Do not wait until copyediting. At the same time, add factual sources to the source log and image or table details to the asset register. If AI output is rejected, mark it as not used rather than leaving the status unclear.

Before formatting, reconcile the manuscript and assets with the log. Mark generated material as included, replaced, or removed; do not reclassify it as human-created merely because it was extensively rewritten. Independently authored replacement content needs its own origin record. Verify or remove unsupported factual claims and keep any optional author note consistent with the actual process.

Before uploading to KDP, compare your final PDF, EPUB, cover file, and metadata draft against your records. Then answer KDP’s AI disclosure questions according to the current wording in your account. If you are unsure how Amazon classifies a particular use, review the current KDP Help pages and choose the most accurate, conservative answer for the files you are publishing.