Introduction to GenAI and Novels
First and foremost, I’m a novelist. Four of my novels were published by Imajin Books and one translated into German and published by Luzifer Verlag. I’m also a professional editor and have helped hundreds of authors improve their stories. I love the process of writing and editing a novel and care deeply about stories and authors.
But what gives me a unique insight in this discussion is that I also have a computer mathematics degree and run a tech company called Fictionary. I love technology and all the ways it can help people be more productive and creative.
Like many of you, I’ve been reading obsessively about generative AI (GenAI) and whether it’s good or bad for fiction writers. The computer scientist in me is thrilled by all the leaps forward in generative AI—it’s literally changing and improving every day—but, if I’m honest, the writer in me is a bit terrified. What will this mean to me and my fellow authors who have spent years working on our craft? Will there still be a place for us if the market is saturated by books written in seconds by artificial intelligence?
I’m trying to stay open-minded. I love how technology can help writers tell their stories, and I am trying to view this new tech as “Intelligent Assistance” for creative people rather than artificial intelligence.
Going forward, some writers will create their novel using 100% human intelligence. Other writers will create their novel using 100% GenAI content. Some of us will be somewhere in the middle of this scale.
We all need to find our place because GenAI is here. It’s up to writers to decide how to use the technology to create powerful stories that the human brain connects with.
In this article, I’m going to look first at all the ways GenAI excites me as a writer. Then I’m going to look at all the ways it makes me feel wary. And then I’m going to show you what happened when I performed a professional edit on a 100% AI-generated novel. I approached it in the same way I perform professional edits on human-generated novels.
Before you read this article, I want you to know where my biases come from.
I wrote this article without using GenAI. The technology I used includes Canva, WordPress, Notion, and ProWritingAid (for grammar checking but not using their rephrase function). I also had the help of other humans (Fictionarians) for ideas, for structure, beta reading, and for proofreading.
I chose not to use GenAI, because I’m writing about how humans can use AI to write novels, and I didn’t want my thought process influenced by AI.
This article is to record our current thinking as we learn and evolve our thinking. Just as technology evolves, our understanding of how to use it productively and ethically will evolve.
What Excites me About GenAI
There are so many things that excite me about GenAI. We are at a tipping point where the way we interact with computers is about to change dramatically. So, to start, I wanted to share some of the ways I think GenAI and fiction can live happily ever after.
Increased Accessibility of Storytelling
In 2015, a YouGov poll in the UK asked people what their dream job would be and a staggering 60% said they would love to be an author. The desire to tell stories is deep inside of us as a species. Imagine all the stories that are trapped in people’s heads because they don’t have the training or time to get them down on paper.
Writing is hard. Ask any author, and they will tell you about the years they spent agonising over every word. Writing is a craft that must be learned. Just as a painter must spend the time learning to mix colours and studying how light hits objects, writers spend years learning the techniques that allow their words to connect with their readers.
I bet a lot of you reading this have had a moment where you had a great idea in your head but when you tried to write it down, it just sounded amateur and embarrassing. Maybe GenAI will help writers get started with that idea and build it into something special.
The idea that GenAI could help more people share their stories is immensely appealing to me.
Generating Ideas and Creativity
Writers who understand strong storytelling are already a step ahead of those whose first step into the writing arena will be to use GenAI. If a writer understands how to outline using story structure, they will be more able to prompt GenAI in a way that is effective.
We all get stuck generating ideas, and GenAI can produce ideas very quickly. An author can use GenAI like a brainstorming partner without the worry of shutting down that partner when ideas are rejected.
The Rise of the Artisan Author
In any supermarket you can get cheap cheese. It’s not amazing, but it does the job of melting on your burger when you need it. This hasn’t meant that people stopped splashing out on Gruyère. Indeed, a whole industry of artisan cheese producers has risen up to offer a higher quality, handmade product.
Maybe human authors need to think of themselves as creating an artisan product. Their dedication to craft and human connection with their readers will allow them to hold on to a loyal fan base, even when there is more market saturation than ever.
Where I Feel Wary of GenAI
Death of New Ideas
GenAI, at its core, is just a serving up new versions of existing texts. It’s scanned millions of texts and learned what words usually follow each other and in what order, and it creates new writing based on that. It doesn’t actually know anything.
Generative AI can mimic but not create art.
I think of human intelligence as having the ability to adapt and learn based on cognitive skills. We learn through experience and problem solve based on that experience. AI learns from data and can use that data to mimic humans.
Imagine what our literary landscape would be without the following authors’ contribution.
Without Dante (1265-1321), we wouldn’t have terza rima. First and third lines end in rhyming words in the first stanza. This is where the last word of the second line in the first stanza rhymes with the last word on the first and third lines of the second stanza.
Jane Austen (1775-1817) invented “free indirect style”. Third person narrative combined with first person speech wouldn’t be a tool used to writing fiction. This was the start of what we call POV writing. She used the style to convey dialogue and thought. In Emma, she only allowed a character to know things experienced (sights, taste, touch, hear, smell). Austen merged narration with character, leading the way to point of view and close point of view that is popular in commercial fiction.
Leo Tolstoy (1828-1910) was the first to introduce interior monologue into narrative fiction. This meant a reader could experience what a character was thinking and feeling, bringing the reader and character closer together. Prior to Tolstoy, writers summarized character’s thoughts and emotions.
These are just a few examples of the ways the human authors changed storytelling in incredible ways and were replicated in a thousand different ways.
Authors Untrained in Story Structure
To get GenAI to outline a novel that had a strong story structure, we had to tell GenAI what the story arc scenes must do. Then we provided a blurb and asked it to create an outline. It took us several tries to to get an outline that followed a strong story structure.
Without knowing what strong story structure is, an author could easily create a poor story.
GenAI also created an outline where the climax scene did not address the story goal. Creating a goal at the beginning of a story is essential. An author needs to make their reader care about whether the protagonist reaches it or not. Will the detective find out who murdered the librarian? Will brokenhearted Sofie find true love? Those are the stakes at play. If you have a climax that doesn’t address that essential goal, then your reader will feel cheated.
An author must have a strong understanding of what makes a good story in order to use GenAI effectively to write the story. This may change in the future as the technology evolves, but for now, the writer must still know their craft.
Ethical Concerns Around Using Author’s Names in Prompts
I’ve heard from many authors that they are uncomfortable with the idea that the AI generators are trained based on the writings of human authors. An author may take years to develop their own style and voice and the technology is using their work without their consent is akin to theft.
I’m still working my way through the murky ethical waters here. On one hand, writers have always taken inspiration from the writers that came before them. You can clearly see the techniques and styles of Stephen King in many (most?) of the horror novels that have been written over the last couple of decades.
Yet I recognise that this is different. Particularly if writers use specific authors’ names to prompt their story. If I tell GenAI to write a story in the style of Dan Brown, then I am acting unethically. Dan Brown, whether you are a fan or not, has a very clear authorial voice. He spent years developing his style and technique and has been incredibly successful doing so. Dan Brown always follows the Story Arc to a tee, but his voice and style are what makes his stories original.
So, I always recommend that if you are going to depend on GenAI to help you write a story, you must start by building out your narrative using trusted structures like the story arc rather than using a specific author’s name.
When I was first playing around with GenAI, I asked it to write a scene for a thriller novel. I was impressed with the output, up until I one of the main characters in the story’s name rang familiar. Robert Langdon. Remember him?
I pulled the old copy of The Divinci Code by Dan Brown from my bookshelf and there, on the pages, was the same scene that GenAI created for me. Sure, the words had been changed to synonyms and some things had been added or deleted, but it was clear that this had been the basis for the scene generated.
Now, I have an uncanny ability to remember names in fiction (much more than my ability to remember names in real life!) and so I recognized it. But if you were trying to write a story using genAI and didn’t recognize this scene or the character name, there is a good chance you could get in trouble for accidental plagiarism. Since GenAI doesn’t offer sources, there is almost no way to check.
For me, that’s a serious red flag.
Over-saturation of the Market with AI-generated Stories
I’ve heard from many authors that they worry the market will become over saturated with AI-generated stories. I’m less worried about this. Great books will rise up and bad books won’t get bought.
Hemingway famously said “The first draft of anything is sh*t”. It’s turns out that’s true for both human-generated first drafts and (currently) AI-generated first drafts.
Even if you are not a novelist, you know that it’s hard to write anything perfectly the first time. But the first draft is an essential step in the process. You need to write the draft before you can start making it better.
Once you’ve written the first draft, that draft needs a story edit because if there is no structure, then there is no story, meaning there is no book.
Let’s look at the story arc.
Time has proven many commercially successful stories follow the form of the story arc. This means to attract a wide audience, a story fits the way the human brain loves to experience a story, because the bulk of the population won’t relate to an unconventionally told story.
Beowulf and Twilight: Two stories written 2758 years apart. Two stories about monsters. Two stories that are unique. What do Beowulf, an epic poem written circa 750BC, and Twilight, written in 2008, have in common? Both are about monsters, but what’s extremely interesting is the both follow the same story arc.
At Fictionary, we use natural language processing to help authors compare the structure of their story to that same structure so that they can tell if they are hitting the right beats at the right time.
When I’m using technology, I make creative decisions on whether I want to follow the story arc or not. If I chose not to, I’m doing it from a place of knowledge, and I’m making an artistic decision. I’m not doing it by accident.
Now you know the importance of the story arc, so let’s move on to specifics.
Story Editing a GenAI Book: Problems I Found
We performed both a story edit and a copyedit on GenAI novels. Below are the results.
Story Structure Issues Create by GenAI
For an author to use GenAI successfully, they must understand the role of each story arc scene and where that story arc scene should be in the story. The simplest form of the story is made up of the inciting incident, plot point 1, the middle plot point, plot point 2, and the climax. The illustration below shows where the scenes should occur in a story.
Story arc scenes in the wrong place
During the first story edit we performed, we found that GenAI placed plot point 1 in the scene immediately following the inciting incident. This means the story lacked depth and felt rushed.
The scenes between the inciting incident were missing. There was no reaction to the inciting incident, no resistance to the story goal, and no lead up to plot point 2. For a writer to use GenAI, they must know these scenes are required if they want readers engaged in the story.
Story arc scenes not performing the correct function
In one GenAI book we edited, the climax scenes did not answer the story goal. In every story, the protagonist has a goal they are trying to achieve. The climax scene must show the protagonist either achieving their story goal or failing to achieve the goal. If it doesn’t, the story isn’t over.
Story arc scenes were too short
The story arc scenes are the most important scenes in a story and the word count per scene is critical to evaluating if the story gives enough page time to these scenes. In the word count per scene shown below, the inciting incident is scene six. The word count for scene three overshadows scene six and lessens its importance.
These three examples have shown an author must understand story structure if they are going to use GenAI to write a novel.
Character, Plot, and Setting Issues Created by GenAI
The next level of story structure to assess are the characters, plot and settings story elements for every scene. Here we found plot holes, repetition, characters not in motion, scene lost in time and space, lack of entry and exit hooks, and boring scene ending types.
Plot Holes and Inconsistencies
GenAI memory is an issue. We found glaring plot holes where a character is in one place in a scene, and the next scene they were in another. This may sound ok, but not when it was impossible for the character to get to the next location. I suspect this links back to the story arc. If key scenes are missing from the story, the story doesn’t make sense.
We used GenAI to create a story outline. Here’s a specific example of a plot hole in the outlining including the prompts we entered:
Fictionary Prompts: I spent a little more time with your second outline. There is a timeline issue. between the opening image and the lead up to the inciting incident. Can you see it or shall I let you know?
GenAI Answers: Apologies for any timeline issue in the outline. I appreciate your attention to detail. Please go ahead and let me know about the specific timeline issue you’ve noticed between the opening image and the lead up to the inciting incident, and I’ll do my best to address it and make the necessary adjustments.
Fictionary Prompts: How can Jaz be a widow when the husband’s body is found in the next scene?
GenAI Answers: You are absolutely right, and I apologize for the oversight in the timeline. It seems there was a mistake in the placement of events in the outline. Here’s an updated version that corrects the timeline issue:
Another plot inconsistency we found for our novel was for a character with agoraphobia. GenAI quickly had this character out and about in a busy town. This character would never do that. A reader would get to this point, find the story unbelievable, and put the book down. This points to GenAI’s inability to be human.
Repetition (sometimes of entire scenes)
On a high-level structural scale, GenAI had trouble remembering the main event in every scene. This means it repeated the essence of one scene a few scenes later in the story.
On a smaller scale, GenAI repeated opening types.
- Scene 1 opened with: The icy wind nipped at her cheeks as she stood alone amidst a sea of black coats in the cemetery.
- Scene 2 opened with: The biting chill of the winter night gnawed through the walls of Jaz Cooper’s house, piercing the tranquility of her solitude.
These two are too close to be interesting to the reader.
There were scenes where the characters were not in motion and created a talking head syndrome. Entire scenes were written almost fully in dialogue. In a novel, this means the scenes lack depth and the reader will skim the scene.
Scenes not Anchored in Location and Time
Readers don’t like to get lost. This issue relates back to the plot holes. When a reader doesn’t know where a character is or the timing of the scene related to the previous scene, they get frustrated and stop reading.
Scene Entry and Exit Hooks
GenAI had trouble generating a strong entry hook either because the POV goal for a scene was unclear, the scene didn’t flow from the previous scene, or a plot hole caused confusion.
Without strong entry hooks in every scene, a reader will stop reading.
Scene Ending Types
The scenes tended to favor ending a scene with thought. There was a lack of variety in scene ending types that could bore a reader.
Copyediting a GenAI Book: Problems I Found
I must admit that the GenAI text was pretty clean when it came to grammar, spelling, and punctuation, which is so helpful. But there is a lot more to good writing at the sentence level than just good grammar. The editing tool ProWritingAid offers a lot more depth when it comes to copy editing. Let’s look at some of the ways that the GenAI text still needed work.
Repeated Words and Phrases
Overuse of unusual words. This comes back to the same issue I mentioned above around memory.
The scene GenAI created for us used the word mystery four times in a 1049 word scene.
- His persistent stare was an unwelcome distraction, his presence an unsolved mystery.
- The funeral was no longer just a platform to bid farewell to her beloved husband, it was now a platform of unanswered questions and unsettling mystery.
- The weight of his gaze seemed heavier, his intent clearer yet shrouded in mystery.
- Her husband’s funeral was supposed to be a sorrowful goodbye, not a dread-filled mystery
The word fear was used four times.
- Fear nestled in the pit of her stomach, spreading its icy tendrils through her veins.
- Her husband’s funeral had now become a battlefield, a war between her fear and her determination to honor her husband’s memory.
- But she knew one thing – she wasn’t going to stand by and let her fear consume her.
- She fled from the cemetery, from the mourners, from the stranger, from her fear.
It’s too much. Your reader will notice the repetition and it will ring as untrue or amateur.
Copy editing 101 warns authors against using a ton of unusual dialogue tags. New writers often use them to try to add colour or impact to an exchange, but it usually just ends up sounding amateur.
Let’s look at a couple of examples.
“Sounds like something he’d do,” Bob observed. The problem is linking the dialogue tag observed to the dialogue. How does a character observe something he is saying? The sentence doesn’t make sense.
“How could I possibly know?” Bob replied warily. There are two problems here. Bob asked a question, so replied doesn’t work. He’s not replying to someone. The addition of the adverb warily weakens the dialogue.
For this, I uploaded an original text:
My family closed in on me and sweat peppered the back of my neck. I loved them. They were good to me. There were just so many of them.
The rewritten option recommended:
My family’s presence was so overwhelming. I felt the sweat gather on my neck. I loved them; they had been kind to me, but there were just too many of them.
A good copyeditor would recommend changing the text GenAI recommended.
My family’s presence was
so overwhelming. I felt the sweat gather Sweat gathered on my neck. I loved them; they had been kind to me, but there were just too many of them.
Two problems with I loved them:
- There is an antecedent issue. The recommended text is saying I loved the sweat and my neck.
- The use of semicolons is not recommended in fiction.
The second sentence recommended by the rephrase option was:
My family huddled around me and I felt a bead of perspiration trickle down my neck. They meant the world to me and were always so kind to me. It was just that there were so many of them.
Notice the addition of
- Filtering: I felt a bead of perspiration. A copyeditor would recommend removing the filtering caused by the word felt.
- Extra word: and were always so kind to me. A copyeditor would recommend removing the ‘so’ and possibly using a stronger word than kind.
- Extra words: It was just that**.** A copyeditor would recommend removing ‘It was just that’.
Now let’s look at examples taken from scenes generated by GenAI.
Both GenAI recommendations strip the author’s voice from the text, slow the pacing of the story, and distance the reader from the character.
Story Editing and Copyediting Summary
These were short examples to illustrate an author using AI to generate their story is still responsible for story editing and copyediting that story. It’s not much different from a human writing a story in the sense that a draft created using AI and a draft created without using AI needs a story edit and a copyedit.
Conclusion: Generative AI and What it Means for the Writer
There is no question that Generative AI is going to change writing forever. There will be some people who will attempt to use it to replace human author voices, others who will use it to enhance human author voices, and others who don’t use it at all.
I believe there is incredible potential for using it as an intelligent assistant to help authors write and edit more effectively.
We are still not at the point where GenAI is going to put out a strongly structured, polished story on its own. Tools like Fictionary play an essential role helping writers turn their imperfect first drafts (whether human-generated or AI-generated) into a story that will resonate with readers.
Fictionary is here to help artists leverage technology to create more stories readers love. Regardless of how the story was created, we will help make sure the story is a story readers love.
Technology can take care of the tasks required to write, edit, and publish a book. The ability to create original stories in a unique voice remains with the writer.