An Originality-First Model for Digital Publishing
A New Kind of Publishing Proposal — Catherine Lea, October 2026
1. AI is creating a new challenge for publishers: how do they distinguish genuinely original work from the growing volume of material that feels familiar, formulaic or machine-generated?
If the aim of banning AI in authorship is to stem the flow of AI-generated query letters, I would understand. But my interest lies elsewhere. What if publishers considered a different model: accepting submissions of AI-written stories that offer something genuinely different, publishing selected titles digitally, and letting readers decide whether they succeed or fail?
A glut of samey books might produce negative responses, although some readers may be perfectly happy with familiar stories. Either way, wouldn't it be worth testing the work in the market rather than dismissing it simply because AI was involved?
2. The idea: a different publishing model
The proposal is to create an alternative publishing route for AI-written and AI-assisted stories, rather than automatically rejecting them because of how they were produced.
The key question would not simply be whether a book was written by a human or by AI. It would be whether the story offered something distinctive enough to deserve publication.
Publishers could accept submissions for a digital imprint specifically designed to find unusual, original or differently told stories. Selected works would be published digitally, allowing readers to respond to them in the marketplace.
The idea is experimental. Some books might fail. Some might succeed. A glut of repetitive books might attract negative responses, although some readers may be perfectly happy with familiar stories and recurring formulas.
Instead of trying to predict every outcome in advance, publishers could test the work and learn from readers.
3. Using AI to identify originality
AI could potentially assist with the initial assessment of submissions.
It might help identify:
Premises, plots and themes that have become overused.
Stories that repeat familiar formulas without adding anything new.
Fresh treatments of familiar subjects.
Unusual combinations of ideas.
Distinctive ways of handling information and revealing a story.
Voices, structures and storytelling choices that stand apart from the crowd.
Human editors would still be needed to assess the strongest submissions and make the final decisions.
The intention would not be to use AI to produce yet more books. It would be to investigate whether AI could help publishers identify the books that offer something different.
There would also need to be safeguards. An assessment system designed primarily to predict commercial success could reject unconventional or challenging work. The aim should be to discover originality, not merely to identify what resembles an existing bestseller.
4. The importance of human experience
Human writers bring a wide range of experiences and perspectives to their work.
Someone who has climbed a mountain, raised a disabled child, lived through an extraordinary event or developed a particular understanding of the world may bring something distinctive to a story. An imaginative writer may also explore possibilities that have no direct equivalent in ordinary experience, including entirely different dimensions or realities.
AI can help a writer develop ideas and shape material, but the writer's experience, judgement and creative choices can contribute to the finished work in ways that matter.
At the same time, lived experience alone does not guarantee good writing. Human writers can produce derivative work, while writers using AI can create something genuinely original.
The question should therefore be what the finished story achieves, rather than making assumptions about its quality from the label attached to it.
5. Publishing as a market experiment
A digital imprint could offer a relatively low-cost way to test selected titles.
Publishers could assess results using measures such as:
Sales and revenue.
Reader reviews.
Completion rates, where available.
Repeat purchases.
Whether readers go on to explore other books from the imprint.
No single measure would tell the whole story. A book might sell modestly but attract devoted readers, while another might generate initial curiosity without encouraging readers to return.
The important point is that the publisher would be testing the work with actual readers rather than relying entirely on assumptions about what readers will accept.
A controlled experiment could also help establish whether different kinds of AI involvement produce different results.
6. The wider problem with publishing
The existing publishing system is not perfect. Editors and agents receive large numbers of submissions, and potentially worthwhile work can be overlooked.
Simply rejecting all AI-written fiction would remove one category of material from consideration without necessarily solving the problem of repetitive or unoriginal writing more broadly.
An originality-first digital imprint could be a way to investigate a different approach. It would not replace conventional publishing or human authors. It would create another route through which stories could be assessed and tested.
The proposal is not that AI-generated fiction should automatically be published. It is that publishers should consider whether some of it might be worth publishing, and whether technology could help them identify the work most deserving of attention.
7. Working concept note
An Originality-First Model for Digital Publishing
Traditional publishing is struggling with two challenges: an overwhelming volume of submissions and growing concerns about AI-generated fiction flooding the market with repetitive material.
Rather than simply rejecting AI-generated work, publishers could establish a digital imprint that uses AI to help identify manuscripts offering something genuinely different.
The system would assess submissions for familiar or overused premises, repetitive themes, originality of treatment, distinctive storytelling and the freshness with which information is handled. AI would assist with initial screening, while human editors would evaluate promising work.
Selected titles would be published digitally as part of a controlled market trial. Reader engagement, completion rates, reviews, repeat purchases and sales would help determine which works merit further investment.
The aim would not be to replace human writers or editorial judgement. It would be to test whether AI-assisted screening can help publishers discover original work that conventional submission processes might overlook.
The guiding principle is simple: instead of producing more of the same, use technology to help identify what is different, then let readers help determine what succeeds.
8. How to take the idea forward
Possible next steps discussed included developing a concise proposal and approaching people or organisations with an interest in publishing innovation.
Potential routes included:
Digital publishing or innovation teams within publishing houses.
Publishing trade publications.
The Publishers Association of New Zealand.
Publishers interested in running a limited digital experiment.
The proposal would be strongest if presented as a constructive publishing experiment rather than simply a criticism of current AI policies.
A working title might be The Originality Imprint or Beyond the Bestseller Formula.
9. Publishing the idea publicly
One option would be to publish a developed essay on Catherine's own author website, authorcatherinelea.com, and use a shorter Facebook post to direct interested readers to it.
Keeping dated drafts and notes would help document the development of the proposal. A copyright notice could identify the author of the essay.
Publishing the essay would establish a public record of the idea as Catherine articulated it. However, copyright generally protects the original wording of an essay, rather than giving someone exclusive ownership of a broad publishing model or business idea.
10. Intellectual property and the conversation with ChatGPT
Discussing the proposal with ChatGPT does not, by itself, mean that Catherine has surrendered her rights to her own writing or given up the ability to develop and publish the proposal under her name.
There is an important distinction between the original expression of an idea and the idea itself.
New Zealand's Intellectual Property Office explains that copyright protects original expression, not ideas, information or methods as such.
OpenAI's Terms of Use, effective 1 January 2026, state that, as between the user and OpenAI and to the extent permitted by law, the user retains ownership of their input and owns the output. The terms also note that outputs may not be unique.
Practical steps include retaining the conversation and dated drafts, publishing the essay under Catherine's name if she chooses, and seeking legal advice if the proposal develops into a commercial venture requiring more specific protection.
11. The central proposition
The proposal is not about choosing AI over human writers.
It is about asking whether publishers could use AI to help find original work, regardless of the assumptions people might make about the method of writing.
A transparent digital publishing experiment could combine technological screening, human editorial judgement and real reader responses.
Some books would fail. Some might succeed. The results could help publishers understand what readers actually want, rather than relying exclusively on predictions.
The aim is not to manufacture more of the same. It is to find what is different, publish it, and let readers help determine what works.
— Catherine Lea
A note on authorship: I used AI to help develop and articulate this proposal, but the original idea and the thinking behind it are mine. And yes, I'm perfectly comfortable saying that. You're welcome.
2. The idea: a different publishing model
The proposal is to create an alternative publishing route for AI-written and AI-assisted stories, rather than automatically rejecting them because of how they were produced.
The key question would not simply be whether a book was written by a human or by AI. It would be whether the story offered something distinctive enough to deserve publication.
Publishers could accept submissions for a digital imprint specifically designed to find unusual, original or differently told stories. Selected works would be published digitally, allowing readers to respond to them in the marketplace.
The idea is experimental. Some books might fail. Some might succeed. A glut of repetitive books might attract negative responses, although some readers may be perfectly happy with familiar stories and recurring formulas.
Instead of trying to predict every outcome in advance, publishers could test the work and learn from readers.
3. Using AI to identify originality
AI could potentially assist with the initial assessment of submissions.
It might help identify:
Premises, plots and themes that have become overused.
Stories that repeat familiar formulas without adding anything new.
Fresh treatments of familiar subjects.
Unusual combinations of ideas.
Distinctive ways of handling information and revealing a story.
Voices, structures and storytelling choices that stand apart from the crowd.
Human editors would still be needed to assess the strongest submissions and make the final decisions.
The intention would not be to use AI to produce yet more books. It would be to investigate whether AI could help publishers identify the books that offer something different.
There would also need to be safeguards. An assessment system designed primarily to predict commercial success could reject unconventional or challenging work. The aim should be to discover originality, not merely to identify what resembles an existing bestseller.
4. The importance of human experience
Human writers bring a wide range of experiences and perspectives to their work.
Someone who has climbed a mountain, raised a disabled child, lived through an extraordinary event or developed a particular understanding of the world may bring something distinctive to a story. An imaginative writer may also explore possibilities that have no direct equivalent in ordinary experience, including entirely different dimensions or realities.
AI can help a writer develop ideas and shape material, but the writer's experience, judgement and creative choices can contribute to the finished work in ways that matter.
At the same time, lived experience alone does not guarantee good writing. Human writers can produce derivative work, while writers using AI can create something genuinely original.
The question should therefore be what the finished story achieves, rather than making assumptions about its quality from the label attached to it.
5. Publishing as a market experiment
A digital imprint could offer a relatively low-cost way to test selected titles.
Publishers could assess results using measures such as:
Sales and revenue.
Reader reviews.
Completion rates, where available.
Repeat purchases.
Whether readers go on to explore other books from the imprint.
No single measure would tell the whole story. A book might sell modestly but attract devoted readers, while another might generate initial curiosity without encouraging readers to return.
The important point is that the publisher would be testing the work with actual readers rather than relying entirely on assumptions about what readers will accept.
A controlled experiment could also help establish whether different kinds of AI involvement produce different results.
6. The wider problem with publishing
The existing publishing system is not perfect. Editors and agents receive large numbers of submissions, and potentially worthwhile work can be overlooked.
Simply rejecting all AI-written fiction would remove one category of material from consideration without necessarily solving the problem of repetitive or unoriginal writing more broadly.
An originality-first digital imprint could be a way to investigate a different approach. It would not replace conventional publishing or human authors. It would create another route through which stories could be assessed and tested.
The proposal is not that AI-generated fiction should automatically be published. It is that publishers should consider whether some of it might be worth publishing, and whether technology could help them identify the work most deserving of attention.
7. Working concept note
An Originality-First Model for Digital Publishing
Traditional publishing is struggling with two challenges: an overwhelming volume of submissions and growing concerns about AI-generated fiction flooding the market with repetitive material.
Rather than simply rejecting AI-generated work, publishers could establish a digital imprint that uses AI to help identify manuscripts offering something genuinely different.
The system would assess submissions for familiar or overused premises, repetitive themes, originality of treatment, distinctive storytelling and the freshness with which information is handled. AI would assist with initial screening, while human editors would evaluate promising work.
Selected titles would be published digitally as part of a controlled market trial. Reader engagement, completion rates, reviews, repeat purchases and sales would help determine which works merit further investment.
The aim would not be to replace human writers or editorial judgement. It would be to test whether AI-assisted screening can help publishers discover original work that conventional submission processes might overlook.
The guiding principle is simple: instead of producing more of the same, use technology to help identify what is different, then let readers help determine what succeeds.
8. How to take the idea forward
Possible next steps discussed included developing a concise proposal and approaching people or organisations with an interest in publishing innovation.
Potential routes included:
Digital publishing or innovation teams within publishing houses.
Publishing trade publications.
The Publishers Association of New Zealand.
Publishers interested in running a limited digital experiment.
The proposal would be strongest if presented as a constructive publishing experiment rather than simply a criticism of current AI policies.
A working title might be The Originality Imprint or Beyond the Bestseller Formula.
9. Publishing the idea publicly
One option would be to publish a developed essay on Catherine's own author website, authorcatherinelea.com, and use a shorter Facebook post to direct interested readers to it.
Keeping dated drafts and notes would help document the development of the proposal. A copyright notice could identify the author of the essay.
Publishing the essay would establish a public record of the idea as Catherine articulated it. However, copyright generally protects the original wording of an essay, rather than giving someone exclusive ownership of a broad publishing model or business idea.
10. Intellectual property and the conversation with ChatGPT
Discussing the proposal with ChatGPT does not, by itself, mean that Catherine has surrendered her rights to her own writing or given up the ability to develop and publish the proposal under her name.
There is an important distinction between the original expression of an idea and the idea itself.
New Zealand's Intellectual Property Office explains that copyright protects original expression, not ideas, information or methods as such.
OpenAI's Terms of Use, effective 1 January 2026, state that, as between the user and OpenAI and to the extent permitted by law, the user retains ownership of their input and owns the output. The terms also note that outputs may not be unique.
Practical steps include retaining the conversation and dated drafts, publishing the essay under Catherine's name if she chooses, and seeking legal advice if the proposal develops into a commercial venture requiring more specific protection.
11. The central proposition
The proposal is not about choosing AI over human writers.
It is about asking whether publishers could use AI to help find original work, regardless of the assumptions people might make about the method of writing.
A transparent digital publishing experiment could combine technological screening, human editorial judgement and real reader responses.
Some books would fail. Some might succeed. The results could help publishers understand what readers actually want, rather than relying exclusively on predictions.
The aim is not to manufacture more of the same. It is to find what is different, publish it, and let readers help determine what works.
— Catherine Lea
A note on authorship: I used AI to help develop and articulate this proposal, but the original idea and the thinking behind it are mine. And yes, I'm perfectly comfortable saying that. You're welcome.
Comments