Sunny Greenhill – How to Write a Books with ChatGPT (страница 8)

This is not a document for court. It is a document for your own clarity and for those cases when partners want to understand your role.

How to save dialogues with AI, and whether you even need to

Here beginning authors often torment themselves: "Do I have to save the entire chat?"

If you are working for yourself and without contracts – not necessarily. But if you are planning:

a publisher,

rights sales,

a serious commercial publication,

and especially a book where the topic of AI is part of the concept,

then it is useful to save at least key fragments: prompts that determined the structure, style, requirements, and the moments where you made decisions.

The simplest way is not to keep everything, but to make "snapshots" of the process:

copied a successful prompt and answer into a separate "Process" file;

added a couple of lines about why you chose exactly that;

moved on.

That way you preserve not chatter, but checkpoints.

Why versioning is your best friend

I know many people do not like versions. It seems as if they interfere with the flow of creativity.

But versioning is not about bureaucracy, it is about the author's power over the text. It is a way to see how a book becomes a book.

And in work with AI, versioning is especially important, because AI tends to "give a lot all at once." And authorial work is mainly about "knowing how to choose and rework."

You should have a habit: do not overwrite the old version until you are sure the new one is better.

This, by the way, is useful not only legally, but artistically as well.

What to do if you have already written the book without documentation

This is also a common situation. A person has already made the manuscript, and then reads chapters like this and thinks: "Damn, I do not have anything left."

That is unpleasant, but not fatal.

There are two practical steps:

1. Create an "after-the-fact" process document.

Describe how you worked: the structure, the stages, what you did yourself, where you used AI.

It is not ideal, but it is better than emptiness.

2. Take the book through deliberate editing and preserve a trace now.

Treat the editing as authorial work: rebuild, align the voice, strengthen the key points.

Save the "before/after" versions.

As a result, even if the early traces are gone, you will gain a provable layer of human contribution.

Practical "legal hygiene" for an author who writes with AI

I want you to remember this not as a set of rules, but as a behavioral habit:

You do not have to be a lawyer.

You do have to be an adult author who can show what they did.

Rights become "clean" not because you said the correct phrase, but because your process looks like an author's process: there is a concept, there is structure, there is reworking, there are decisions, there is a trace.

Conclusion of section 1.6

You do not need to turn writing into an archive. But you do need to leave enough "living fingerprints" so it is clear: the book was not simply generated – it was created.

If you document:

the concept,

the structure,

2-3 versions by chapter,

and short decision notes,

then you gain both artistic clarity and legal stability.

1.7. Publication and Platforms: What Matters So You Do Not Run Into Blocks, Rejections, and an "Unclear Status" for the Book

You can write a wonderful book… and still run into an unpleasant wall at the final step – publication. And often the problem is not even the text itself, but the uncertainty around it.

So when you come with a book in which AI participated, you enter an environment where people have two questions in mind:

is it safe to publish?

who bears responsibility and who owns the rights?

If the platform does not have a clear answer, it acts like any protective system: it either asks follow-up questions or plays it safe.

I want you to understand this in advance. Then you will not treat the requirements as "censorship" or "unfairness." You will see that this is normal risk management.

Why platforms get nervous about AI

Let us begin honestly: the problem is not that "AI is bad." The problem is that with AI, the amount of the following has sharply increased:

mass-produced junk content made "on a conveyor belt";

books assembled from other people's texts and retellings;

unoriginal compilations;

pseudo-guides where everything sounds confident but is factually wrong;

materials that violate someone else's rights because the author did not check them.

Platforms see this not from one case, but from thousands. And they develop a reflex: if content looks "automatically produced," they begin to review it more strictly.

This may irritate you, but it is better to accept it as reality: AI is a trigger for heightened scrutiny.

And that means your task is not to prove that "AI is fine," but to show that your book is not junk generation, but authorial work.

What most often causes problems in publication

There are three typical situations where beginners break down.

The first is "unclear authorship."

This is when wording appears in the metadata, on the cover, or in the description that looks like co-authorship with AI. This immediately creates legal and editorial anxiety.