GUEST POST by Jeremy Esekow
A few months ago, my company, Storywise, ran a writing competition that felt slightly reckless, but completely necessary. We pitted humans against AI storytelling.
I came into this from two very personal places.
First, as an author. I’ve sent out countless query letters over the years. Most were met with silence. Second, as an entrepreneur and employer. I’ve also sat behind the other side of that inbox, staring at hundreds of job applications I simply didn’t have the capacity to answer. I know what it feels like to wait. And I know what it feels like to be overwhelmed.
And lately, another question has been nagging at me.
If AI can now generate entire books in minutes, what happens when the number of submissions to agents and publishers multiplies overnight? What happens when slush piles triple… or grow tenfold? Will it become dramatically harder for human authors to stand out? Will truly original voices be buried under a tidal wave of competent, machine-generated slop?
The Humans vs. AI Storytelling Contest
So instead of wondering, we decided to test it. Rolling up our sleeves, we built a website entry page and set the rules for our writing competition (including a strict no-AI-assistance policy for the authors). Then we assembled judges, secured prizes (first prize was a publishing contract), designed ads, and got banned by reddit (obviously). Then we opened the floodgates!
The rules were simple. Each author submitted one entirely human written chapter from a new manuscript. We then took the logline from the authors chapter, fed it into an AI writing tool, and generated an AI chapter.
Our human judges were tasked with two jobs:
- For each submission, decide whether the author wrote a better chapter than the AI generated version. We gave the judges specific guidelines.
- And from there, choose which were the three best written chapters submitted overall.
What followed was both reassuring and unsettling.
We received genuinely wonderful entries. Many chapters were brave and distinctive, feeling undeniably human. In the end, the prizes went to three fully human authors, leaving me deeply proud. And relieved.
But there were moments when the AI chapter held its own. When they were compelling, mimicking structure and pacing convincingly. And if AI can produce something “good enough” at scale, and submissions grow exponentially, then the challenge for human authors may become insurmountable.
After everything wrapped up, we asked our judges what they had learned. Had the experiment changed their view of AI writing? Were they more optimistic? More concerned?
Their answers were thoughtful and nuanced. Here are the five biggest takeaways from the competition:
1) Human dialogue is better than AI dialogue.
AI-generated dialogue often lacks human-like emotional nuance and authenticity. Readers could sense that it is “artificial dialogue”, and it consequently creates distance instead of genuine character interaction.
The AI generated chapters consistently underperformed in scenes requiring dialogue. Whether crafting distinctive and believable messaging, original intonations, or mannerisms, (a good!) human author seemed to deliver the goods better.
2) AI is very bad at pacing.
We actually struggled to get an LLM to write even one chapter longer than 2,000 words. And when it did succeed lengthwise, it generally told most of the story, leaving little content for the rest of the book.
AI struggles with pacing. It’s just too rushed. It can’t make things play out slowly. AI is great when you need to move things along, but as soon as slow burn is required, it collapses. When shocking readers with a horrific ‘bear eating camper’ scene, a good author will have you sweating for pages, not just a paragraph.
3) Author voice needs to come through from the get-go.
One thing the judges kept returning to was voice. Again and again, they found themselves drawn to manuscripts that carried a distinctive, unmistakable authorial voice. We’ve seen this pattern before.
Many of Storywise’s customers, when setting up their taste profiles, don’t just describe genres or themes. They describe voices. Sharp, lyrical, commanding, comforting, or intimate. They want to feel the person behind the page. That truth became even clearer in a competition where we were literally comparing human writing to AI-generated prose.
AI storytelling can produce structure and coherence. It can even produce elegance. But I don’t believe it can produce a voice that feels owned and truly fresh. In several cases, the judges chose the human chapter even when it was less polished or less “publisher-ready.”
4) Did the editor want to read on with AI storytelling?
The strongest predictor of a high score was surprisingly simple: did the judge want to keep reading and to know what happens next? Interestingly, authors who tried to pack too much into chapter one often scored lower.
Overcrowded openings with too many characters, too many events, too much backstory left judges feeling overwhelmed rather than intrigued.
Some chapters had the opposite problem. They opened brilliantly, with tension and promise, but then wrapped things up a little too neatly by the final page. Judges found themselves enjoying the chapter without feeling compelled to turn the page. Others lingered too long in scene-setting with beautifully drawn worlds but weren’t able to pull the judge into a compelling narrative. Delivering an intriguing opening or creating an unmissable hook at the closing were great ways to leave our judges begging for more.
5) AI storytelling is getting very good
At times, it was genuinely difficult to tell which chapter was written by a human and which by AI. We even ran blind tests, presenting the two versions side by side without labels. In more than a few cases, judges guessed wrong. That alone was sobering.
For the anti-AI hardliner, there are clues. One telltale sign of machine-generated prose is over-describing details that don’t carry any narrative significance. Confident writers won’t describe the color and quantity of flowers stuffed into a vase in a corridor that the protagonists briefly passed through. If it’s not important to the story, a good author will know not to waste ink and reader attention on it. (Obviously the AI hasn’t been taught about Chekhov’s Gun...ARA)
And then there’s the EM-dash. AI adores it. Find me a ChatGPT response that doesn’t include an Em-dash. Sadly, authors are starting to shy away from the EM-dash so as not to have their writing labelled as AI generated.
The bottom line though is that AI is getting very good at creative writing. Not independently and not magically. Only when it’s being directed by a competent storyteller. Good prompters know exactly what they are doing and how to get the most out of AI assistance and editing tools.
Humans Won vs. AI Storytelling, but it was Close
To sum it all up, we ran our first writing competition. Humans won in the end. It was very close. Buy me a beer and I’ll give you the uncensored version. And we learned five clear lessons that you, as an author, can take with you, as you continue to thrive amidst rapidly changing dynamics:
- Human dialogue is better than AI-generated dialogue. Invest in your dialogue and give your characters distinctive voices.
- AI is terrible at pacing. It can’t do slow burn and left to its own devices, rushes through the plot. Master the art of sustaining tension. Learn to draw readers forward without resolving everything too soon and you’ll not only endure in the age of AI, you’ll thrive.
- Judges love a distinctive authors voice, especially when they are being swamped with AI-generated ‘slop’. Focus on delivering content with the best version of your own unique style.
- Keep your reader wanting more with a great hook. Get the reader invested in your characters right out of the blocks, or build towards a heart-stopping cliffhanger. Otherwise, you risk your readers not wanting to turn to the next chapter.
- And finally, we must acknowledge that AI is getting very good at creative writing. It still has blind spots. It struggles with true originality, lived emotional complexity, and the unexpected twists that make a story unforgettable. But as a tool for researching, drafting, refining, and editing, it is improving rapidly.
This post was written by Dr. Jeremy Esekow (@StorywiseNet) March 1, 2026
***
What about you, scriveners? Have you ever pitted your deathless prose against AI storytelling? Have you read any books or stories generated by AI that you didn’t recognize right away as AI slop? Do you have any questions for Jeremy?
PLEASE NOTE: This is a guest post written by Dr. Jeremey Esekow, co-founder of Storywise (bio below). It was NOT written by Anne R. Allen, who is NOT employed by Storywise, had nothing to do with this contest, is not South African, and does not have a PhD. Most blogs host guest posts. For some reason, a lot of people are having trouble understanding the concept. I got 5 pitches from authors who wanted to write about AI this month. I thought Dr. Esekow’s post was the most balanced and interesting. Sending Anne poison pen notes does not change reality. Nasty, unhinged comments go to spam. Don’t try to prove robots are better than humans at understanding English!
Dr. Jeremy Esekow is the co-founder of Storywise. He’s a South African author who earned his master’s degree in business and PhD in philosophy from the University of Cape Town. He’s worked in the high-tech sector for the past fifteen years, most recently co-founding Storywise, a software platform for publishers, agents and authors. Prior to that, he worked in tourism and spent time in huts and hotels all over Southern Africa. You can also find Jeremy on LinkedIn
Storywise was founded in 2023 to address a challenge familiar to writers and publishers alike: giving promising authors a better shot at publishing successfully and providing publishers and agents with a better way to discover stories that fit the wishlist.
***
featured Image: Mohamed Nohassi for Unsplash
Thank you, Jeffrey and Anne, for a fascinating but unsettling article. The contest yielded great info for us humans.
Jeffrey, I’m curious what you think about agents and editors using AI to screen submissions. That would seem to streamline the process and give faster responses to authors. But I also wonder what happens when AI screeners start judging AI-written submissions. Will they favor AI b/c it sounds like them?
The slope grows slipperier every day.
Hi Debbie,
See my reply below
Anne, I really appreciate you and Jeremy testing what AI can actually do today.
One gentle observation: I don’t think the future is “human vs AI” so much as pure AI pushed by companies trying to cut margin versus human + AI, where authors use the tools as part of a structured craft process. I’ve sat in meetings with top levels of big publishing companies, unless there are humans willing to step up and steer this, it’s too cost effective to just let the engineers figure this out.
One-shot prompting is usually the weakest test. On Friday we open-sourced the workflow we’re using on our YouTube channel (Future Fiction Academy): 17 steps—story dossier → plan → draft → three editing passes → finalize → process summary—to keep narrative logic intact and improve things like dialogue over revisions.
Thank you for doing careful, reality-based work here. It matters.
Hello Elizabeth,
Many thanks for your comment.
I wholeheartedly agree with you. I’ve heard it said several times to literary professionals that AI should be viewed as a hand, not a brain.
I’m going to check that 17 step process out!
Ouch! The contest results are a bit too close for comfort, said the concerned—yet still emdashing—author.
Yes – right!
Thanks for your comment
While working with my publisher on a book cover, we discovered huge amounts of AI-generated artwork. Artists might be more endangered than authors.
Thanks for your comment Fred!
I think you are right.
I read that AI generated images make up more than half of adobe’s 600 million image stock library
A writer friend with well over a dozen titles to her mystery series, tested AI by asking it to “write a chapter in the style” of her author name. After it scoured its vast database of stolen, er, “scraped” books, it produced a chapter with the name of her protagonist and the household cat reversed, among other obvious gaffes. Still, the results were creepily similar.r
HI Gay. Thanks for the anecdote!
I’ve seen those gaffes so many times! Admittedly, less and less as time passes.
This is such an important post. Thank you for these insights. I am of the mind that AI-generated stories (wholly or partially) will continue to muddy and distort the book industry. I am now seeing authors make statements on their copyright pages and in submission letters that their work is “human-authored” to stand apart from the onslaught of AI-generated fiction. The honor system can work and hopefully isolate AI-generated fiction. The Authors Guild has human-authored seals now to clearly identify to readers, book buyers, and all in the editorial arm of this industry that their book was not produced by AI bots. I think the more authors and writers who declare their work authentically created by a human and not a bot, the better for all in the industry and for readers.
Hello Paula,
Thank you for your thought provoking comments.
I’ve seen a few platforms that authenticate writing as human. It’s an important development for readers who are focused on consuming 100% human creativity. I think more nuance will be needed and have many thoughts on how to tackle this.
Anecdotally, I wrote a linkedin post a while back, comparing two separate instances where an author forgot to remove prompt instructions from the published versions of their book. In the sci-fi author’s case, readers rebelled. With the romance novel, the book continued to sell just as well.
Hi Anne,
Great, great post! Thank you for bringing this to us.
As an editor, I’ve been wearing hip-waders for AI research, trying to arm myself against AI-produced material presented as human-produced. Information like this is a tremendous help in giving me a more balanced view.
In the end (as everyone seems to say), my opinion, which I use to keep the even keel, is that AI is incapable of creating original work, it must be ‘fed’ human-written creation (see blood-sucking of published works by AI) in order to regurgitate any words. This fact should be pounded into the front line of all AI discussion/support/defense/etc. It doesn’t matter what it is eventually used for, by humans, it is still an ‘it’, not a living thing with a life experience from which to draw creative writing.
In the past year, I have been astounded at the number of ‘authors’ who, when questioned about their use of AI, begrudgingly admit they ‘just used AI for structure, grammar, punctuation, some story content, maybe a bit of plot options and character or world-building’…and often explain the use this way: “I’m not a very good writer, but this is a great story!”
Thanks again for this very revealing and informative post.
Many thanks for your insightful comments Maria.
We’ve seen many authors on our author services platform praise AI as a tool for research and tightening up.
For many years prior, authors they had been using AI powered platforms like Grammarly and google for the same things, without pushback. The LLMs have really just supercharged these abilities.
Future debates will most likely be centered on how much AI assistance readers are prepared to tolerate in the actual fabric of the story.
Wonderful post, and terrific test! Thank you for this. My publisher contracts state that I can’t us AI for any part or purpose, and I’m good with this. But I can’t help being sad. Most writers already make far less per hour of work than publisher staff and bookstore staff. It seems we are close to being devalued even more.
Thanks for your comments Melodie,
Indeed. It does feel like author’s income is shrinking and I’ve seen recent research that bears this out.
Just curious, does your contract allow you to use AI tools for line editing (e.g. Grammarly)
What a great idea to have a human/AI writing contest. The judging info and takeaways are valuable—should be interesting to see what’s next. (And yes, I inserted that em-dash all by myself.)
Thanks for your comment Kay,
We enjoyed the competition and plan to run a second later this year – perhaps around September :).
Thank you for your comments and questions Debbie.
We learned what you just pointed out the hard way. AI likes its own writing and rates it highly (obviously – I guess). It also seems to have no feel for authentic or engaging writing.
For these reasons, we annoyed all of our users by trying to judge whether manuscripts are good or bad.
Our company, Storywise helps agents and editors prioritize the most relevant submissions. We base our assessment not on writing quality, but rather on a comparison between the manuscript themes, characters etc….. and the wishlist that the editors/agents set up. I think that this is the most effective use of AI in the submissions process
A very interesting article, thanks. I assume the competition entrants knew what was going on, and agreed in advance to their entries being fed into an AI tool for the purposes of this review? I mention this because not only was there the initial uproar about authors having their works pirated to populate AI, but now there are instances of the general public feeding new original work into AI without permission, in the (often) misguided attempt to “check” if it’s AI generated!
Hi Clare, thanks for your comments!
We were absolutely clear and transparent regarding all terms, conditions and the judging process.
More than that, we host our own AI infrastructure and protect against leakage of author content into the publicly available AI models. Storywise places a very high value on privacy and copyright protection, while at the same time enjoying the incredible efficiencies that AI offers.
Such an interesting post. I love that Jeremy conducted this experiment. And the results were fascinating. There are still skills that only humans can master. Who knows when this could change, but for now I think we should embrace our humanness and write from our unique human hearts and heads. Thank you, Anne!
Many thanks Christine,
I couldn’t have said it better!
Wow! That’s a really interesting contest. Good to know that humans still have an edge. I know I’d rather read an average book by a fellow indie author than the best book AI could ever generate. (Plus, humans use less water and other resources. 😉)
Great post, excellent guest, wonderful insight.
“In the quiet of the night, online friendship speaks the loudest, as we share our stories and bond over our experiences.” ― medicosaurabh (Goodreads user)
J (he/him 👨🏽 or 🧑🏽 they/them) @JLenniDorner ~ Speculative Fiction & Reference Author and Co-host of the April Blogging #AtoZChallenge international blog hop
Thanks for your comments J!
Agreed. Nothing beats a lived experience.
I can’t imagine a machine expressing the essence of human spirit and soul. Is it possible that the machines are approaching the ability mimic or surpass believability ?Can they express, for example , the love and connection a child would have for its doll or teddy bear ?
Carl–In the contest described in the post, AI lost, and the humans won every time. But you bring up an interesting point. Young humans do develop affection for inanimate objects. And so, we’ve discovered this week, do other primates, like Punch the lonely baby monkey who became attached to the plush monkey. So maybe it’s just one step for adult humans to develop affection for inanimate objects fueled by AI. I keep reading about people who “fall in love” with AI chatbots. I think there lies madness, but maybe, for people feeling like Punch the monkey, fake love is better than none.