I use AI to write. So why might I refuse to read an AI-generated book?

I care about the final product, but books and music complicate that position. AI can help build a book. It shouldn’t replace the human reason for the book existing.

I use AI heavily in my own writing. It helps me develop arguments, make connections and turn unfinished thoughts into work I can publish.

Yet I’m not sure I would choose to read an AI-generated novel. I might feel the same about an AI-generated album, even if somebody told me it was technically good.

That creates an uncomfortable question. Why should somebody value my AI-assisted work if I’m reluctant to consume somebody else’s?

The easiest defence would be that Editorial Intelligence is business writing rather than art. Its purpose is to explore ideas, test frameworks and make professional experience useful. Nobody is coming to it expecting the equivalent of a novel or record.

But that distinction is a little too convenient. People could still look at the amount I publish and wonder whether the thinking is mine or whether ChatGPT is manufacturing an intellectual identity for me.

What makes the work mine?

My answer cannot simply be that I edit the copy or approve the final version.

The value I claim to bring to Editorial Intelligence comes from what exists before the draft: 15 years of journalism and content work, the problems I encounter at Sage, experiments with AI, reactions to things I see on LinkedIn and the connections I make between them.

AI does not independently experience any of that. It helps me interrogate the material, organise it and express it more quickly. Sometimes it also challenges my initial position and helps me find the more interesting argument underneath it.

But I decide what is worth exploring. I provide the context. I reject arguments that don’t feel true. I remain responsible for what gets published.

That distinction only holds while I continue doing those things. If I began asking AI to generate articles about Editorial Intelligence and publishing whatever came back, the work might retain the same surface style while losing the reason anyone should care about it.

The question is therefore not whether AI touched the writing. It is where the thinking and important decisions came from.

A book would require more from me

I could imagine using AI to help write a book. In fact, given the way I now work, it would be strange to pretend I wouldn’t.

For a nonfiction book, AI could help me organise research, test the argument, identify repetition and develop early drafts. That would be an extension of the Editorial Intelligence process I already use.

But a book would require much more substantial editing and rewriting. Someone has to spend several hours inside it. The voice, rhythm and structure would need to feel consistently mine rather than like a competent reconstruction of my ideas.

Even significant editing might not be enough on its own. If AI supplied the premise, argument and structure while I merely improved the prose, I would be closer to the editor of a synthetic manuscript than its author.

I would need the original reason for the book to come from me. The central argument, experience and intellectual structure would need to be mine. I would have to make the important narrative decisions and be able to explain why every part belonged there.

Fiction would create an even more difficult boundary. If AI generated the characters, plot and emotional moments, rewriting the sentences might not make the imaginative act mine. I could use it to question a scene, test alternatives or help solve structural problems, but I would still want the world and story to originate substantially from a human imagination.

The process can be part of the product

I still believe the final product matters. Bad work does not become good because a person suffered while creating it. Human effort is not an automatic mark of quality.

But books and music do something different from most business content. When I read a novel or listen to an album, I’m not only consuming effective combinations of words or sounds. I’m encountering another person’s imagination, experiences, taste and decisions.

An AI-generated song might convincingly reproduce emotional expression without anybody having experienced the emotion. A novel might contain believable characters without its supposed author ever having imagined their lives beyond a prompt.

The finished work could still affect me. I’m not prepared to declare it worthless simply because of how it was produced.

But I might still choose not to give it my attention. There are already more books and records made by people than I could experience in a lifetime. Wanting some evidence of another human mind behind the work is a reasonable preference, even if an AI-generated alternative could pass a blind quality test.

Cinema makes the boundary harder

The argument becomes more complicated when I think about film, because cinema has always manufactured reality.

Sets are constructed. Performances are edited. Miniatures become cities. Computer-generated creatures interact with actors who never saw them. A CGI dinosaur is not real, but its appearance, movement, lighting and role in the scene are still the result of thousands of human decisions.

So “real versus generated” is not a useful boundary.

Martin Scorsese recently made that particularly visible by using Black Forest Labs’ FLUX image generator to create storyboards for What Happens at Night. His stated purpose is to communicate the image in his head more clearly to the people making the film. He isn’t asking AI to write the screenplay or generate the finished movie.

Steven Spielberg has drawn his own line elsewhere. He has said he can accept practical uses such as location scouting, but not AI making decisions about writing, characters or storytelling.

Their positions suggest a distinction between AI executing a human intention and AI supplying the intention itself. Scorsese describing a scene he has already imagined and using AI to visualise it is different from asking a model to invent a Scorsese-like scene.

But even that distinction is not enough.

Storyboard artists have objected because translating a director’s words into images is itself creative work. Scorsese may preserve his own authorship while removing somebody else’s judgement from the process.

This means I am actually dealing with several boundaries at once.

There is creative agency: who originated the idea and made the consequential choices?

There is expressive purpose: was AI helping somebody communicate something, or manufacturing something to fill a cultural content slot?

There is provenance: what work was used to build the system, and did those creators consent?

And there is labour: is the technology extending somebody’s capabilities, or removing collaborators whose interpretation previously shaped the result?

That labour question applies to my own work too. It isn’t enough for me to say the Editorial Intelligence ideas originate with me if the wider adoption of this way of working also changes what happens to editors, researchers, designers and junior writers. Individual authorship and collective consequences are different questions.

There may not be one line

Creative work has never been the untouched expression of one person. Books have editors. Records have producers, session musicians, samples, synthesisers and software. Films are profoundly collaborative constructions. AI could become another instrument within all of those processes.

“Made with AI” will eventually tell us very little.

The more revealing question is where a piece of work sits on a spectrum. A person might use AI to execute a specific intention. They might explore possibilities with it. AI might begin proposing important creative decisions. Eventually, the person may do little more than select and approve what the machine produced.

At some point, “author” starts becoming “commissioner”, “curator” or “editor”. But the transition cannot be measured by the percentage of words, notes or pixels generated. It depends on where the originating idea, judgement and responsibility sit.

Did somebody have something they wanted to express? Did they make the important creative decisions? Is the finished work recognisably theirs because they shaped it, rather than because a model successfully imitated them? And whose creative contribution disappeared in order to make that process faster?

That applies to my work too.

I don’t think using AI disqualifies somebody from calling themselves a writer, musician, filmmaker or artist. But using AI myself does not oblige me to value every possible use of it.

My emerging boundary is that I’m open to work created by people using AI. I’m much less interested in work generated to simulate the presence of a person who was barely involved.

AI can help build a book. It shouldn’t replace the human reason for the book existing. But even that does not settle everything, because a genuine human intention can still be executed through tools that remove other human creators from the process.

Perhaps the boundary is not between real and generated at all. It is between AI executing human intention and AI supplying the intention, judgement and expression we thought we were encountering — while remaining honest about who may have been displaced along the way.