News··8 min read
Does the EU AI Act require authors to label AI-assisted books?
The EU AI Act’s labeling rules have applied since August 2, 2026. What they mean for novels, nonfiction, AI covers and authors outside Europe, plus a checklist.
On this page
For most novels, no. Since August 2, 2026, the EU AI Act’s transparency rules have applied, but the duty to label AI-generated text covers text published to inform the public on matters of public interest, and even then not when a person has reviewed it and takes editorial responsibility. Realistic AI images of real people or places are a separate case. Here is what the rules say, who they apply to, and what a careful indie author should do now.
What changed on August 2, 2026
The EU AI Act (Regulation (EU) 2024/1689) came in stages. Its transparency rules in Article 50 became enforceable on August 2, 2026. They are about honesty toward readers, viewers and listeners: people should know when content was made or changed by AI in situations where it matters.
In June 2026 the European Commission also published a final Code of Practice on marking and labeling AI-generated content. It is voluntary, but companies that sign it can use it to show they comply.
Who has to do what: AI tools versus authors
The Act splits the work between two groups.
- Providers, meaning the companies that build AI systems, must make synthetic text, images, audio and video “marked in a machine-readable format and detectable as artificially generated or manipulated”. This is a technical watermark or metadata job for the tool maker.
- Deployers, meaning anyone who uses an AI system in a professional capacity, have the visible labeling duties. Article 3 excludes purely personal, non-professional use. Writing and selling a book is professional activity, so a self-published author using AI is a deployer.
So the question for you is not whether your book carries a hidden watermark. It is whether anything you publish needs a visible notice.
Novels and other fiction
Article 50(4) asks deployers to disclose AI-generated or manipulated text only when it is “published with the purpose of informing the public on matters of public interest”. A thriller, a romance or a fantasy series is not published to inform the public about public affairs, so the plain reading is that fiction text falls outside this duty. Law firms that have written about the rule read it the same way, though no court has ruled on books yet.
There is a second safety valve. Even for public-interest text, no label is needed when the content “has undergone a process of human review or editorial control” and a person or company “holds editorial responsibility for the publication”. An author who drafts with AI, rewrites, checks and signs the book as its author is doing exactly that.
Nonfiction on health, politics, science and other public topics
This is where you need to be careful. Commentators list topics such as politics, public health, security, the environment, consumer safety and science as typical matters of public interest. A book on vaccine history, an election explainer or a climate guide could fall inside the rule if it contains AI-generated text.
The fix is the same human-review exception. If you review and edit the AI-assisted text yourself, check the facts, and you, or your publishing imprint, take responsibility for the book, the labeling duty does not apply. What the rule targets is AI text pushed out to the public with nobody standing behind it.
- Keep notes of how you used AI and what you checked, in case anyone ever asks.
- Put your name, or your imprint, clearly on the copyright page as the party responsible for the content.
- If a section really was published largely as the AI wrote it, say so in a short note.
AI cover images and illustrations
Images follow a different rule. A deployer must disclose a “deep fake”: AI-generated or manipulated image, audio or video that resembles existing persons, objects, places, entities or events and would falsely appear authentic. A painted-style fantasy dragon or a stylized silhouette is not a deep fake. A photorealistic picture of a real celebrity, or a real city scene that never happened, could be.
For an evidently artistic, creative or fictional work, which a book cover usually is, the Act limits the duty to disclosing that such content exists “in an appropriate manner that does not hamper the display or enjoyment of the work”. In practice, a line on the copyright page such as “Cover image created with AI tools” is a sensible way to do it. You do not need a sticker on the front cover.
Authors outside the EU who sell in Europe
If you live in the United States, Canada, the UK or Australia, the rules can still reach you. Article 2(1)(c) applies the Act to providers and deployers in third countries “where the output produced by the AI system is used in the Union”. A book sold on Amazon.de, Amazon.fr or Kobo in the EU is output used in the Union.
In practice that means the same advice as above: fiction is mostly outside the text duty, review and take responsibility for nonfiction, and add a short note for realistic AI images of real people or places. Fines for breaking the transparency rules can reach EUR 15 million or 3% of worldwide turnover, and for small businesses whichever amount is lower applies. That figure is aimed at platforms far larger than an indie author, but it shows the rules are meant seriously.
How this fits with Amazon KDP’s own AI rules
The EU Act is not the only rule you face. Amazon’s KDP content guidelines ask you to tell Amazon about AI-generated text, images or translations when you publish or republish a book, even if you edited them heavily afterwards. AI-assisted work, where you wrote the content and used AI to edit, refine or brainstorm, does not need to be reported to KDP.
The two rules ask different questions. KDP wants to know how the content was made, for its own records. The EU Act cares about whether readers are misled on public-interest topics or by realistic fakes. Answer KDP honestly in the publishing form, and handle the EU side with the human review and copyright-page note described here. Our guide on how to use AI disclosure on KDP walks through the form.
A practical checklist for 2026
- Write and revise the book yourself, using AI as an assistant, and read every line before you publish.
- For nonfiction on public-interest topics, check every fact against a real source and keep notes of your checks.
- Name yourself, or your imprint, on the copyright page as the party responsible for the content. Our free copyright page generator can draft one.
- If the cover or interior uses photorealistic AI images of real people, places or events, add a short line such as “Cover image created with AI tools”.
- Answer the KDP AI questions honestly for every new edition.
- Keep an eye on guidance from the EU AI Office, which announced task forces on these rules for providers and for deployers from September 2026.
Frequently asked questions
Do I have to write “made with AI” on my novel’s cover in the EU?
No. The text-labeling duty covers text published to inform the public on matters of public interest, which a novel normally is not. A cover image only needs a disclosure if it is a realistic deep fake of real people, places or events, and for a creative work a short note on the copyright page is enough.
Does the EU AI Act apply to self-published authors in the US or UK?
It can. Article 2(1)(c) covers deployers outside the EU when the AI system’s output is used in the EU, for example a book sold on Amazon.de or Amazon.fr. The duties are the same narrow ones described in this article.
Is AI-assisted editing covered by the rules?
The provider’s marking duty does not apply to AI that only provides standard editing assistance or does not substantially change the input. For authors, text you wrote and then polished with AI is your own text, and if you review and take responsibility for it, the deployer labeling duty does not apply.
Does this replace the Amazon KDP AI disclosure?
No. KDP’s question is a separate, Amazon-wide rule. Disclose AI-generated text, images and translations to KDP when you publish, whatever the EU rules say.
Sources
- EU AI Act, Article 50: Transparency obligations (applies from August 2, 2026) — checked 19 November 2026
- EU AI Act, Article 2: Scope (third-country providers and deployers) — checked 19 November 2026
- EU AI Act, Article 3: Definitions (deployer, deep fake) — checked 19 November 2026
- EU AI Act, Article 99: Penalties — checked 19 November 2026
- Regulation (EU) 2024/1689 on EUR-Lex — checked 19 November 2026
- Cooley: EU AI Act transparency obligations take effect August 2, 2026 — checked 19 November 2026
- Addleshaw Goddard: Code of Practice on marking and labeling AI-generated content (June 2026) — checked 19 November 2026
- Amazon KDP content guidelines (AI-generated content) — checked 19 November 2026
This article explains the rules in plain language. It is not legal advice. If your book deals with sensitive public-interest topics, check with a lawyer in your country.