AI Transparency Policy

KDP AI Studio uses large language models and image-prompt models supplied by third-party providers to assist your publishing workflow. When you run a feature, your prompt and the relevant project context are sent to the model provider, and the returned output is stored in your workspace.

• Research and trend analysis: demand, competition and opportunity signals summarised by a model.
• Metadata generation: titles, subtitles, descriptions, keywords and category suggestions.
• Structure and content: outlines, interior page plans and image prompts.
• Optimisation tools: SEO review, quality checks and cover concepts.

Every AI action shows its credit cost before you confirm it. Data handling for prompts and outputs is described in our Privacy Policy.

AI models predict plausible text; they do not know facts. You should expect and plan for the following limitations:

• Hallucination: confident statements, statistics, citations or market figures that are simply wrong.
• Stale knowledge: model training data has a cutoff and may miss recent marketplace or policy changes.
• Non-determinism: the same prompt can produce materially different results.
• Bias: outputs can reflect biases present in training data.
• Similarity: common phrasing may resemble existing published titles or descriptions.

Research signals, demand scores and competition estimates are directional guidance produced by a model. They are not measured marketplace data and should not be treated as such.

You are the author and publisher of everything you produce with KDP AI Studio, and you are responsible for it once you publish. Before publishing you should:

• Read and edit every output rather than publishing it unchanged.
• Confirm the content complies with Amazon KDP's content guidelines and AI-disclosure requirements, and with those of any other marketplace you use.
• Ensure the work does not infringe third-party rights — see our DMCA Copyright Policy.
• Follow our Acceptable Use Policy at all times.

Verify any factual, historical, medical, legal, financial or safety-related claim against an authoritative source before publishing. Run generated text through your own plagiarism and quality checks, validate keywords and categories against the live marketplace, and review image prompts for trademarked or recognisable elements. Our built-in quality tools assist this review — they do not replace it.

Our commitments in how we build and operate AI features:

• We do not use your private project content to train public AI models.
• We apply safety filters and enforce our acceptable use rules against harmful generation.
• We disclose that outputs are AI-generated inside the product and in this policy.
• We select providers with published safety and data-handling practices, and we minimise the data we send them.
• We support human review of any enforcement decision that affects your account.

Models, prompts and safety systems change over time as providers release new versions and as we improve our own tooling. This means output quality and style may shift between releases. We monitor aggregate quality metrics, act on user feedback and update this policy when our approach materially changes; the "Last updated" date above always reflects the current version.

AI output is provided "as is" without any warranty of accuracy, originality, completeness, fitness for a particular purpose or non-infringement. We make no guarantee of marketplace approval, sales, revenue, ranking or reader reception. To the extent permitted by law, liability arising from your use of AI output is limited as set out in our Terms of Service. Questions? Email support@kdpaistudio.app.