Takeaway Summary about AI and literature:
- AI can imitate literary style, but lacks long-term narrative memory and metaphorical consistency.
- AI outputs may reflec cultural memory of authors as much as their original works.
- AI disrupts biographical readings, prompting a return to text-focused criticism.
- Authorship in AI-assisted literature demands active human curation, editing, and intention.
- AI democratizes creative production, but poses pedagogical challenges and opportunities.
- Students and educators must learn to evaluate and work critically with AI tools.
- BARD409 integrates queer theory and literary criticism into its retelling of Othello.
Many people are trepidatious about AI and literature. When I first dreamt up BARD409, it was autumn 2024 in Japan. I’d just sat through a weekend faculty meeting and was decompressing with a walk by the canal beneath the turning leaves. I’d been experimenting with AI to remix creative voices, and suddenly I had the idea of turning Shakespeare into novels by famous writers. But whose voice could I use for such a task? It would need to be a great writer! So, what if, say, Charles Dickens rewrote Othello as a novel? My intention was not for parody or pastiche, this would be a serious stylistic experiment in voice, form, and interpretation.
I imagined the idea would catch fire. Instead, I sat on it. Hesitated. Worried. AI, even in literary form, was (and still is) radioactive in some circles. The backlash can be brutal. The micropublisher I set up, Hungry Wolf Press, has been quite open in its use of AI for experimentation, and even before AI, I was using computers and machines to generate Flarf poetry. One of our other authors shared a critique video on Reddit and was viciously attacked—despite clearly not being the original creator. People don’t read context. They lash out. I’ve heard of artists receiving death-threats for using AI, despite many uses being quite legitimate and AI being baked into tools like Adobe Photoshop and even MS Word, so it’s almost impossible not to use it at this point!
Yet, despite the vitriol, there’s a growing network of researchers and artists exploring AI with nuance. I recently attended a fascinating event at the University of Leeds titled Investigating Human and AI Creativity and Imitation, organised by Dr. Mel Evans. It brought together people working across digital humanities, computation, and literature. The conversations reminded me that using computers for creative purposes isn’t new. These ideas stretch back to the early days of digital computing [click here for a related post about the history of machines and creative writing]. In fact, my earlier project, Moloch, was very much part of that lineage—blending poetry and computation, corpora and machine translation to produce original works. I’m glad I wrote Moloch when I did, as it would not be the same if I wrote it now, it serves as a time-capsule to the limited capabilities of computers at the time, and that was only 2016! It also helped prepare me for the messier, more layered work of BARD409.
Dr. Emily Middleton, a Dickens expert at the University of Leeds, was one of the first people I spoke to in depth about BARD409. She has a background in corpus linguistics and uses AI in her research and teaching, which gives her a uniquely balanced view. We met over Zoom and this piece is a reflective summary of that discussion, paraphrased from my notes and impressions.
Dickens, Shakespeare, and the Machine
Dr. Middleton approached my Shaken-up Shakespeare project with thoughtful curiosity. She was especially drawn to what the model “believed” Dickensian style to be: the flurries of metaphor, strange analogies, and shifting epithets. The fun, as she saw it, was in watching where it landed close to the mark—and where it lost the thread entirely.
She noted how AI struggles to sustain metaphor. Dickens might compare Iago to a cat and Roderigo to a fish and then draw that imagery through entire chapters. The AI? It introduces the metaphor, then forgets it or replaces it. What this shows is how the model is prioritising stylistic markers without deeper structural coherence. This is where the editor comes in. For example, I love the line from Macbeth “O, full of scorpions is my mind”, mainly because I find it weird that a medieval Scottish king would know what a scorpion is! So I wrote a backstory to that line and threaded it all through the novel.
More broadly, we observed that AI outputs are shaped not only by original texts but also by secondary commentary: film adaptations, fan interpretations, lesson plans, even Wikipedia. In a way, the AI isn’t just mimicking the style of these authors. It goes further, actually echoing our collective memory of them. Dickens and Shakespeare are less historical figures than cultural archetypes in this process.
On Authorship, Queerness, and Interpretation
Our conversation quickly turned to authorship. Dr. Middleton raised concerns about how literary culture is increasingly tied to biography. Audiences often demand that writers mirror their characters’ identities. We agreed that AI breaks this expectation. If a text has no clear author, no identity to latch onto, interpretation is necessarily redirected to the work itself.
This became particularly relevant as we discussed my choice to include queer perspectives in Othello by Dickens. I had revisited Shakespeare scholarship on the homoerotic tension between Iago and Othello and chose to make this more explicit. Emily was thoughtful about this addition—recognising its scholarly basis, and also seeing how it allowed the AI-assisted narrative to engage with contemporary criticism.
She also brought up Kenneth Goldsmith, known for his appropriation-based poetry—transcribing traffic reports or entire newspapers and publishing them as art. AI writing shares something of that DNA. It forces us to rethink originality and intent, and to focus instead on language, framing, and effect.
From there, our chat naturally turned to authenticity. What does it even mean in an age where writing is collaborative, recursive, and mediated by machines? We agreed this was a question without easy answers—but a necessary one.
The Work Behind the Work
I was clear with Emily that BARD409 wasn’t just some idle prompting exercise. It took months. I trained multiple models, scrapped entire drafts, edited hallucinated plots, restructured characters, and threaded in critical themes. The AI struggled with narrative memory. One version of Macbeth had the king dying multiple times. Fixing those errors required both technical tweaking and creative restructuring.
Emily was particularly interested in this “conducting” process—the way I guided outputs with large input passages, built prompts that preserved original Shakespearean dialogue, and added critical flourishes. As someone who works with machine learning and corpus analysis, she appreciated just how much steering is involved. It’s not like I pressed magic button and out popped two perfectly formed novels. It’s hard graft, and getting a decent output was quite a struggle. She encouraged me to write more about the process—how much text the AI can handle, where it stumbles, and how editing transforms raw output into something publishable.
AI in the Classroom
We also talked shop. In her own teaching, Emily gets students to critique AI-generated essays and literary analysis. The results are fascinating. The machine might mimic academic voice well enough, but it also misquotes, hallucinates sources, or makes absurd interpretive leaps. This, she said, makes it a perfect teaching tool for provocation. Students who view AI as a shortcut are not the ones getting good grades.
She shared an amusing anecdote: tutors now suspect AI when students use perfect em dashes. The irony being that writing accurately is now suspicious. Tools that improve syntax and structure are changing how teachers evaluate what “effort” looks like.
Translation is another big factor. We’re increasingly seeing students using live translation software to follow seminars and draft essays. AI makes their work possible—but also raises questions about authorship and understanding. Are they learning the material, or just producing acceptable output? These questions are on my mind constantly as well, when I teach students English-taught subjects.
These dilemmas are not going away. But she believes the answer isn’t a blanket prohibition of AI. The answer lies in incorporating these new tools into pedagogy. We need to help students understand what the machine does. Let them question it. Let them test its limits.
Final Reflections
This conversation made it clear that projects like BARD409 sit at the intersection of literary history, critical theory, and digital experimentation. Many people see AI as a threat to literature. But perhaps they’re provocations—ways of asking: What do we value? What do we expect from a text? What happens when authorship becomes plural, entangled, and partially machine-made?
Emily and I both hope to continue this discussion—possibly even in person when we’re next in the same country. I’m deeply grateful for her time, generosity, and insight. Her upcoming book on Dickens’ poetry promises to shed new light on a neglected corner of his work, and I’ll be first in line for a copy.
In the meantime, BARD409 continues to evolve. The next volume may be Romeo and Juliet, though nothing’s certain yet. What is certain is that the questions it raises—the tension between style and substance, the role of authorship, the future of creativity—aren’t going anywhere. Which exactly why we need to keep writing.
