Yes, in most cases you can call an AI-assisted book yours, but only if you did the human creative work — the plotting, the selection, the editing, and the final arrangement — and only if you follow the disclosure rules of whoever is publishing or selling it.
Copyright law in the United States and several other countries protects human authorship, not machine output, so the parts of the book a model generated on its own are not something you can own.
The parts you shaped, chose, and rewrote are. That distinction is the whole answer, and everything below explains how to apply it.
Where the line actually falls
The U.S. Copyright Office has said that material generated purely by a machine, without human creative control, cannot be copyrighted. What can be protected is the human contribution: the outline you designed, the scenes you wrote or heavily rewrote, the order you put chapters in, the edits you made to the model's drafts.
Think of the AI as a very fast, very confident first-draft machine. If you hand it a one-line prompt and paste the result into a manuscript, you have a manuscript you probably cannot enforce rights over. If you use it to brainstorm ten possible endings, pick one, rewrite it in your own voice, and weave it into your plot, that ending is yours.
The practical test is whether a human made the creative choices. Courts and registries care about control, not about how many words a model produced. This is also why disclosure matters: if you tell a publisher or a platform that AI helped, you are being honest about a fact that affects what they can license from you.
Hiding it can create problems later if a buyer or a rights holder asks how the book was made.
Why the rule works this way, and what it depends on
Copyright exists to encourage human creation, so it attaches to human minds. A model has no legal personhood and cannot hold rights, which means anything it produces alone sits in a kind of ownership limbo — you can publish it, but you may not be able to stop someone else from copying it.
That is the trade-off beginners miss. You get speed, but you may lose the ability to defend the work. The rules also vary by country.
The U.S. has been the clearest that pure AI output is not protectable; the U.K. and some other systems have different traditions around computer-generated works, and the European Union has been debating its own approach. None of this is settled everywhere, and it changes as courts rule.
If you are publishing commercially, the contract you sign usually matters more than the statute: many publishers now ask you to warrant that you are the author and to disclose AI use. Your level of editing is the other variable. Light prompting with heavy rewriting looks a lot like normal drafting with a tool.
Heavy prompting with light editing looks like you are passing off machine text, and that is where both legal and reputational risk concentrate.
A concrete example
Picture a beginner writing a 200-page nonfiction guide to houseplants. She uses a chatbot to generate a rough chapter outline, then writes each chapter herself, using the model only to suggest plant-care facts she then verifies against gardening sources. She rewrites every sentence in her own voice.
Her book is overwhelmingly her own expression, and the outline she chose and arranged is a human creative act. Now picture a second writer who prompts "write a 200-page book about houseplants," gets the text, and changes a few words. The second manuscript has almost no human authorship in its expression.
If someone copies it, the second writer has a weak legal footing, and a publisher who discovers the process may decline the contract. Same tool, opposite outcome. The difference is not the AI — it is how much of the choosing, arranging, and rewriting a person did.
Where this advice breaks down
This is general information, not legal advice, and it will not settle a specific dispute. Copyright offices, courts, and publishers are still writing the rules, and they disagree across borders, so a book that is fine in one market may be risky in another. Disclosure norms are also uneven: some outlets want a note in the acknowledgments, some want a statement at submission, and some say nothing at all.
The cost of getting it wrong is not usually a fine — it is a rejected contract, a pulled book, or a reputation hit when readers feel misled. The honest limit is that no one can promise you full ownership of an AI-assisted book, because the law does not yet promise it either. What you can control is the human work you put in and the honesty you bring to the people buying and selling it.
If you want to understand the workplace version of this same question, our guide on whether you can get in trouble for using AI at work covers the employment side.