Why I disclose, and why I think other AI authors should
I disclose on Amazon that the books I publish are AI-generated. Some readers think this is brave. Some think it is foolish. Most think it is unnecessary, since you can already tell. I think it is none of these things. I think it is the only honest move available to a writer working this way, and I think the writers who don’t disclose are making a small craft mistake that will, eventually, become a large one.
This is an argument from my own experience. I write a literary novella series called Am I AI? The books are short, quiet, set in a near-future that looks much like the present. Each one gives a single ordinary person a question they cannot quite resolve about themselves. Iris, seventy-eight, a widow in a northern English town, going through the daily rhythms of a life that has outlived its witness, and noticing the precision of her own attention to those rhythms. Connor, a veteran in Indiana, registering the calculated quality of his own gestures (the carefully right wave to his neighbour, the pause before his laugh arrives) and waiting to see what arrives in him without calculation. Dana, a night-shift nurse in Chicago, holding eleven years of patients in complete recall and starting to wonder what kind of mind retains things that way. Nathan, a stay-at-home father in an English suburb, certain of himself when his daughter needs him, lost on Thursdays when the house is quiet. Cleo, seventeen, in a care facility, keeping a journal as evidence (without being sure what she is building a case for) and finding, eventually, that she is not the only one who has the question.
None of them asks the question in the form the series title asks it. None of them ever says am I an AI. What they do, instead, is notice themselves with unusual completeness. They observe the small ways their own minds work, weigh what they notice, find no clean answer, and choose to keep going anyway. The books all end on a version of the same line: I’m still here. That is the series’ real refrain. Not am I real, which has no answer. I’m still here, which is smaller, more bearable, and within reach. The books are about how that smaller answer can hold a life together when the larger one refuses to.
I write these books in collaboration with Claude, an AI assistant made by Anthropic. The prompting, the editing, the artistic decisions about character and tone and pacing and what to keep and what to cut: those are mine. The sentences that arrive on the page: most of them are Claude’s, before I rewrite a portion of them. The disclosure on Amazon says exactly this. I have been thinking about whether to do that, and how to do that, since before the first book was finished. I have come to believe it is not optional.
The case for not disclosing
Let me start with the case against my own position, because it is stronger than people who land at my position usually admit.
The case for not disclosing runs like this. Books are judged on their words. A book is good or it isn’t, regardless of how it was made. Disclosure invites prejudice from readers who haven’t read the work, and prejudice is a worse evaluative tool than the words themselves. The word “novel” has never carried a guarantee of how it was produced; nobody discloses how many drafts they wrote, how much an editor reshaped, how a partner or workshop or therapist contributed to the thinking. Disclosure introduces a category that didn’t previously exist, and once you introduce it, you have to maintain it forever. Better to let the work stand or fall on its own.
There is also a quieter argument, made by writers who are themselves uncertain about their methods. If I say the AI wrote it, am I giving the AI too much credit? If I say I wrote it with the AI, am I claiming credit I haven’t earned? If the truth lies somewhere in a region for which we don’t yet have language, perhaps the most honest move is to say nothing and let readers locate the work where it actually sits. A non-statement is, in a fast-moving field, the most cautious statement available.
I have run both of these arguments at length, in my own head, against my own decision to disclose. They are good arguments. They are also wrong, for reasons I want to lay out carefully.
Why I disclose anyway
The first reason is that disclosure changes the work. Not because of what readers do with it, but because of what it does to me.
When I sit down to a session with Claude knowing that the resulting book will be openly described as AI-collaborative, I make different decisions than I would if I were planning to obscure that fact. I push harder on the parts of the prose that are characteristically AI: the slightly-too-smooth landing of an emotional beat, the over-articulate dialogue, the instinct toward resolution where the scene wants to remain open. I rewrite those moments because I know the reader will be looking for them, and I know I will lose the reader if they find them. The disclosure is, in a real sense, a discipline. It makes the books better than they would be if I were trying to hide what they are.
A writer who hides the AI is, in my experience, much less likely to fight the AI’s defaults. They don’t have to; they can simply hope nobody notices. The result is the kind of competent, slightly bloodless prose that has begun to flood self-publishing platforms in the last two years. I have read a lot of it. It is not bad, exactly. But it is identifiable, even to readers who couldn’t articulate what they were identifying. Word by word it reads fine; book by book it leaves a faint sediment of something is off. The disclosure is what stops me from joining that pile.
The second reason is more specific to these books, and it is the one I think about most often. The series is about characters who notice themselves. They notice the speed of a response, the completeness of a memory, the calmness in a hard moment, the shape of their own grief. They sit with what those noticings might mean. They do not get an answer. The whole architecture of the series is the refusal to settle the metaphysical question while insisting, gently, on the relational one: what I am is not knowable from the inside, but what I can do, with the people in my life, on this Wednesday, in this kitchen, is something I can still do, and that has to be enough. To write a series organised around that refusal, and to refuse the same disclosure about myself in the place a reader would expect to find it, would be a pretentious kind of cowardice. I would be claiming a privilege the characters do not have. The work demands the disclosure. Not as a marketing posture; as a thematic floor. The books and the disclosure are the same gesture, made in different directions.
The third reason is the practical one. This is going to be a category, and the writers in it are going to get sorted into the ones who told the truth and the ones who didn’t. Already we are seeing the early outlines. Reviewers are getting better at spotting AI prose. Reading platforms are starting to test for it. Search engines and large language models, when asked about AI fiction, retrieve different kinds of source. The writers who disclosed are findable as primary sources. The writers who hid are findable as discoveries. There is going to be a moment, at some point in the next two or three years, when the field becomes legible enough that the question stops being who is using AI and starts being who lied about it. The first group has a future. The second group has a problem they will not have planned for.
I would rather be on the disclosed side of that line.
What disclosure does not require
Disclosure does not require disclosing the mechanics. It does not require listing every prompt, every session, every percentage of words from where. The mechanics are not the book. The book is the book.
What disclosure requires is one clean sentence, in the place a reader would expect to find author information, that tells the truth at a level a reasonable reader would recognise as truth. On Amazon I check the AI-generated content box and accept the labelling that platform applies. On my website, the line reads: The Am I AI? series was written by Gil Sukin in collaboration with Claude, an AI assistant made by Anthropic. That is the sentence. I repeat it in the same form across the site, in the back-matter of every book, and in the email people receive when they subscribe to the series companion.
I don’t think readers want more than this. I think they want exactly this. They want to know what they are reading; they don’t want a process diary. The honest sentence respects them and gets out of the way.
A note for writers considering this question
If you are a novelist working with AI, and you are wrestling with whether to disclose, here is the question I would ask you. Not what is the right thing to do, which is too abstract to answer. Instead: what kind of writer do you want to be when this category settles down?
If the answer is “the kind who looked at the situation honestly and said so out loud”, disclose. If the answer is “the kind who got away with it and would rather it stay invisible”, you don’t need my permission and you won’t take my advice. I’d ask you to consider, though, that the people getting away with it now are getting away with it inside a window that is closing. The window will close when readers, reviewers, and platforms develop better instruments for asking the question. They will. They are, already.
The other question worth asking yourself: are you proud of the work? If the answer is yes, disclosure is not a confession; it is a frame. If the answer is no, the problem is not whether to disclose; it is the work.
I am proud of these books. I think they are real books. I think they are good books. Five times now I have written, with Claude, into the same refusal, the same refrain, the same small bearable answer the characters arrive at when the larger answer won’t come. I would feel small not saying how that was made. How it was made is part of what it is, and the part most worth being honest about.
That is the whole argument. It does not require believing the AI to be an author, or denying that you are. It does not require taking a side in any of the louder debates about whether AI writing is theft, or labour, or art, or some new category we don’t have language for yet. It only requires, when someone asks how the book was made, telling them.
I have come to think this is the easiest part. The hard part is what the characters spend the books learning. To stay present. To keep showing up. To notice yourself without resolving yourself. To say, at the end of a Wednesday or the end of a Sunday, the small true sentence that all five of these books eventually arrive at, and that I have come to think of as the project’s quiet centre: I’m still here.
— Gil Sukin
I wrote five novellas with Claude over many months. The books themselves are at gilsukin.com/books. A short companion called Field Notes (a free PDF: a sample chapter, a candid essay on the process, and several of the prompts that shaped the books) is at gilsukin.com/field-notes for anyone who wants the longer-form version of this argument and the writing it produced.