All posts
AI for Authors

Book Metadata with AI: KDP Rules and Prompts

Book metadata is the information KDP uses to list and find your book: title, subtitle, description, keywords and categories. AI can help you research and draft all of it. KDP requires you to disclose only text, images and translations that an AI tool created, not work you wrote yourself with AI’s help.

SelfPub Go TeamUpdated 12 min read

In short

  • KDP requires disclosure of AI-generated text, images and translations. AI-assisted work you wrote yourself needs none.
  • A published Kindle ebook’s title and subtitle are locked; its description, keywords and categories stay editable.
  • AI drafts often trip the same rules: sales-rank words, invented praise, other authors’ names, repeated keywords and HTML that KDP does not support.
  • Use AI as a researcher that feeds you raw material, and write the final text yourself.
  • Check keywords and categories on Amazon itself: an AI tool can invent phrases and categories that do not exist.
On this page

What counts as book metadata on KDP

KDP’s Metadata Guidelines for Books list the fields: title, subtitle, series, description, author, contributors, publisher, keywords, primary audience, primary marketplace, categories, cover image and ISBN. Five of them decide how readers find and judge a listing, and they are the ones AI drafts touch. The table gives each one’s main rule and whether you can still change it after you publish (Update your book details).

FieldKDP limit or ruleEditable after publishing?Where AI drafts go wrong
TitleOnly the title as it appears on the cover. No sales rank (“bestselling”), promotions (“free”), other authors, trademarks or HTML. KDP’s keyword tips add that customers are more likely to skim past titles over 60 charactersKindle ebook: no. Paperback and hardcover: for 72 hours after first publication“Bestselling” labels, “in the style of” author names
SubtitleTitle and subtitle together under 200 characters; same rules as the titleSame as the titleA subtitle stuffed with keywords
DescriptionUp to 4,000 characters, tags included. Allowed tags: b, em, i, u, br, p, ol, ul, li, h4 to h6. Not accepted: reviews, quotes, URLs, prices, spoilers, emojisYes, at any timeInvented praise, emojis, div and h1 to h3 tags
KeywordsUp to seven keywords or short phrases; the metadata guidelines recommend two or three words. No other authors, sales rank, “free” or program namesYes, at any timeSeven near-duplicates; words already in the title
CategoriesUp to three, from KDP’s list for your audience and marketplace; they must match the bookYes, at any timeCategories picked for reach, or ones that do not exist

KDP’s timelines give 72 hours for an updated ebook description or keywords to appear, and its categories page says a category change can take up to 72 hours to show.

What KDP’s AI policy says

KDP’s content guidelines require you to tell KDP about AI-generated text, images and translations when you publish a new book, and again when you edit and republish an existing one. They separate two cases:

  • AI-generated content — text, images or translations created by an AI-based tool, even if you edited them heavily afterwards. Images include cover and interior art. Must be disclosed.
  • AI-assisted content — you created the content yourself and used AI to edit, refine, error-check or improve it, or to brainstorm and generate ideas. Does not need disclosure.

Disclosure is information you give KDP; the guidelines do not mention any label shown to readers. What gets a book rejected or pulled from sale is different. The same guidelines cover the title, cover art and description, and they do not allow descriptive content meant to mislead customers or that does not accurately represent the book. AI drafts break that rule easily: invented praise, references that do not exist, claims the book does not deliver. Disclosure excuses none of it, and KDP says you are responsible for verifying that AI-generated and AI-assisted content follows all of its guidelines.

Six things in AI drafts that break KDP’s rules

What to strip out before you paste AI output into a KDP listing:

  1. Hallucinated quotes, reviews and rank claims. AI assistants confidently produce phrases like “USA Today bestselling author” or made-up review pulls. KDP accepts no reviews, quotes or testimonials in a description, real or invented, and no promotional material, and its metadata guidelines bar sales-rank references such as “bestselling” from titles and keywords.
  2. Generic descriptive openers. Patterns like “In this gripping page-turner”, “A masterful tale” and “An unforgettable journey” say nothing about your book, and shoppers skim past them. They are not against the rules; they just waste the opening line. KDP’s own advice is to describe the main plot, theme or idea in about 150 words, with a first sentence that draws the reader in (Write a Book Description).
  3. Keyword stuffing. AI tools optimizing for discoverability often suggest seven-keyword sets where several are minor variations of one term, or repeat words already in the title and categories. KDP’s keyword help tells you to skip information covered elsewhere in your metadata, words already in your categories, and variants of spacing, punctuation, capitalization and plurals. It also rules out quality claims such as “best novel ever”, time-sensitive words such as “new” or “on sale”, and Amazon program names such as Kindle+ (formerly Kindle Unlimited) or KDP Select. Each repeat or banned term is a slot you did not use.
  4. Sales language in the description. Lines like “Grab your copy today”, “on sale this week” or “leave a review” do not belong. A description cannot carry price, availability or ordering information, time-sensitive details, advertising, or requests for customer reviews.
  5. HTML and emojis outside the rules. KDP’s description field allows specific HTML — <b>, <em>, <i>, <u>, <br>, <p>, <ol>, <ul>, <li>, <h4>, <h5>, <h6> — and says <h1> to <h3> are not supported (KDP Help). AI tools regularly emit <div>, <span>, inline styles or <h1>–<h3>, and often add emojis, which KDP does not accept in a description either. The limit is 4,000 characters, and the tags count toward it.
  6. Other authors’ names in titles, subtitles or keywords. AI suggestions like “In the style of James Patterson” or “A Brandon Sanderson-inspired epic” are unauthorized references to other authors, which KDP’s metadata guidelines prohibit in the title and the keywords. Use comparable-author phrasing only in promotional material outside the listing itself.

Prompt patterns that produce review-safe metadata

The general principle: treat the AI as a research assistant feeding you raw material, not a copywriter producing finished prose. The metadata you publish should be text you wrote, informed by AI-generated structure and vocabulary, not text the AI generated and you approved.

Title and subtitle

Bad prompt: “Write a title for my cozy mystery about a librarian who solves murders.”

Better prompt: “Here are 30 cozy mystery titles from bestsellers in 2024-2026: [list]. Identify the structural patterns — alliteration, length, name placement, season/theme words — and propose 8 title candidates for my book about a librarian solving murders, with a brief note on which pattern each candidate matches.”

The output is editable input, not finished output. You pick the candidate that resonates with your book’s actual voice and edit it into a title you would have written yourself with more time. Before you commit, run the finalists through the title analyzer to check length, readability and keyword use, and against KDP’s limit: title and subtitle together must be fewer than 200 characters. KDP’s keyword tips add that customers are more likely to skim past titles over 60 characters. The title is also one of the fields you cannot change later (see below), so this is the step to slow down on.

Description

Bad prompt: “Write a 200-word product description for my mystery novel.”

Better prompt: “Read this synopsis: [your synopsis]. Identify the three strongest hooks. List the comparable published novels readers of those hooks already know. Suggest three opening sentences that lead with each hook respectively.”

You then pick a hook, write the description in your voice, and use the comparable-novels list, after checking that every title on it exists, to make sure you have not accidentally over-promised.

Validate the HTML before saving: KDP supports only the tags listed above. In SelfPub Go, the book-launch description editor keeps only the tags KDP accepts. If you are doing this manually, the book description generator gives you a proven structure with KDP-ready HTML.

Keywords

Keywords is where AI assistance is most useful and most commonly misused. KDP lets you enter up to seven keywords or short phrases per book, and its metadata guidelines recommend phrases of two or three words. The trap: filling the seven slots with seven near-identical variations of one core term, which misses the long tail entirely.

Working pattern: ask the AI for “30 search phrases a reader of [book X] might type into Amazon when they don’t remember the title”. Filter to phrases that pass these tests:

  • Two or three words long, as KDP’s metadata guidelines recommend.
  • Not repeating words from your title, subtitle or categories, and no singular and plural twins.
  • Phrases a reader would actually type — natural language, not SEO-style. KDP’s own test: search Amazon for the phrase and read the suggestions in the search box.
  • Setting, character, plot or tone markers, not just genre. KDP’s examples are “single dad”, “coming of age” and “feel-good”.

Leave out quotation marks. Without them, customers can find the book by searching for any of your words; with them, only people who type all of the words do. Pick seven that complement each other. The hard part is the complementing. AI can suggest 30 candidates but cannot tell you which seven balance breadth and specificity for your specific book; that is your judgment call.

The disclosure question, answered specifically

For a metadata workflow that follows the patterns above — AI for ideation, structure and checking, you write the final published text — no disclosure is required. That is AI-assisted content, not AI-generated content. The line is not vague: did you create the text, or did the AI?

If an AI tool writes your title or description and you publish it, even after edits, that text is AI-generated by KDP’s definition, and the guidelines cover the title and description as well as the manuscript. Write the final text yourself, or treat it as AI-generated. The same goes for drafts from any AI tool, including SelfPub Go’s: treat them as raw material for text you write.

For book interior text generated by AI, with or without editing passes, disclose. KDP sets no minimum amount: if an AI-based tool created the text, it is AI-generated even after substantial edits. The same goes for AI-generated cover and interior images, and for translations: a machine translation you then proofread is still AI-generated. Read KDP’s current policy text before each upload; disclosure applies when you publish a new book and again when you edit and republish one.

What you can still fix after you publish

Not every field forgives a mistake. According to KDP’s Update your book details page, a published Kindle ebook’s title, subtitle, edition number and primary author name are locked. For a paperback or hardcover you can edit the title, subtitle and primary author name for 72 hours after the title’s first publication; after that you need a new edition. The description, keywords, categories and reading age stay editable at any time.

That split tells you where to spend care and where to experiment. Take your time over the title and subtitle before launch. After launch, use AI for what stays cheap to change: try a new description, swap a weak keyword, re-check categories. Real search terms are better evidence than any list an AI invents. If you advertise, the search-term report shows what readers typed before they clicked or bought, and our Amazon Ads guide shows how to read it.

What this looks like as a workflow

A working session for new-book metadata:

  1. Comparable books research. Collect 30-50 comparable titles in your subgenre from Amazon’s search results and category pages, with their real descriptions, and spend 20 minutes reading them in raw form. An AI tool can invent titles and descriptions, so paste in the real ones and let it find the patterns.
  2. Title ideation. Generate 8-12 candidates from comp-pattern analysis. Sleep on them. Pick a final.
  3. Description draft. Outline the three beats with AI help; write the description in your voice; AI for editing passes only.
  4. Keyword expansion. Generate 30 candidates; filter manually; test each in Amazon’s search box; pick the final seven.
  5. Category research. AI can suggest category ideas, but only Amazon shows which categories exist in your marketplace and format today, and KDP lets you choose up to three. Picking the right three is your decision, based on where similar books sit. The category finder lets you browse KDP category paths before you choose.

Text you wrote yourself with AI help is AI-assisted under KDP’s guidelines, so it needs no disclosure, and because every claim in it is yours to check, it stays within the metadata rules.

Used like a copywriter, AI can produce listings that get rejected. Used like a researcher, it produces listings that compete. The difference is in how you prompt, and in how much of the final text is genuinely yours.

Frequently asked questions

Do I have to disclose AI use for my book description?

It depends on who wrote the final text. KDP's content guidelines require you to tell KDP about AI-generated text, images and translations, and they apply to the title, cover and description as well as the manuscript. Text created by an AI-based tool counts as AI-generated even after heavy edits. If you wrote the description yourself and used AI to brainstorm, edit or check it, KDP calls that AI-assisted and does not require disclosure.

How many keywords can I use on KDP?

Up to seven keywords or short phrases per book. KDP's metadata guidelines recommend phrases of two or three words. Its keyword page says to skip words already in your title or categories, misspellings, variants of one word (spacing, punctuation, capitals, plurals), quality claims such as “best”, time-sensitive words such as “new”, other authors, quotation marks, and Amazon program names such as Kindle+ and KDP Select. It also suggests testing candidates in Amazon's search box and reading the suggestions it offers.

How long can a KDP book description be?

Up to 4,000 characters, and HTML tags count toward the limit: the word test is 4 characters, and in bold it is 11. KDP supports the tags b, em, i, u, br, p, ol, ul, li and h4 to h6, but not h1 to h3. It does not accept reviews, quotes, URLs, prices, spoilers or emojis. KDP's own advice is to describe the main plot, theme or idea in about 150 words.

How many categories can I choose for my book?

Up to three. KDP offers a list based on your primary audience and marketplace, and the list differs by format and store. Pick categories that describe the book: KDP can remove or change categories that do not match its content. A change can take up to 72 hours to show.

Can I change my title or subtitle after I publish?

Not for a Kindle ebook: its title, subtitle, edition number and primary author are locked once it is published. For a paperback or hardcover you can edit the title, subtitle and primary author for 72 hours after the first publication; after that you need a new edition. The description, keywords, categories and reading age can be changed at any time; KDP's timelines give 72 hours for an updated ebook description or keywords to appear.

Can I use other authors' names in my title, subtitle or keywords?

No. KDP lists unauthorized references to other titles or authors among the items prohibited in a title, and other authors and books by other authors among the items prohibited as keywords. A subtitle follows the same rules as the title. Keep comparisons such as “in the style of” for promotional material outside the listing.

See what each book earns after its ads.

SelfPub Go puts your Amazon Ads spend next to your KDP royalties, book by book, and points out ads that spend without selling.

Create an account