Distant Writing: Literary Production in the Age of Artificial Intelligence
Luciano Floridi1,

(^1) Digital Ethics Center, Yale University, 85 Trumbull Street, CT, 06510, United States
(^2) Department of Legal Studies, University of Bologna, Italy
Abstract
This article introduces the concept of “distant writing”, a novel literary practice in
which authors act as designers, employing Large Language Models (LLMs) to generate
narratives, while retaining creative control through precise prompting and iterative
refinement. Unlike Moretti’s distant reading , which uses computational analysis to
interpret large corpora of existing texts, distant writing harnesses computational tools
(LLMs) to author new narratives, reshaping the literary production process. By
examining theoretical frameworks and practical consequences, and running an
experiment in distant writing called Encounters , this article argues that distant writing
represents a significant evolution in authorship, not replacing but expanding human
creativity within a design paradigm. The distinction between writing (close) and
“wrAIting” (distant) reveals how LLM-assisted creativity can generate narrative
possibilities previously inaccessible, transforming literature’s modal space while
challenging traditional notions of authorship, creativity, and literary production. This
emerging practice merits critical attention as it shapes future literary landscapes and
reconfigures relationships between human creativity and artificial intelligence.
Keywords : distant writing, artificial intelligence, large language models, authorship,
narrative design, modal logic, literary production, wrAIting.

1. Introduction: The Emergence of Distant Writing
Computational approaches to literature have transformed how we analyse and
interpret texts for some time. Franco Moretti’s concept of distant reading^1 introduced
methodologies for examining vast corpora of literature, revealing macro-level patterns
and structures through computational techniques that traditional close reading could
hardly discern or not at all (Moretti 2000, 2013, Kirschenbaum 2016, Piper 2018,
Ramsay 2011). However, while distant reading represents a computational turn in
literary analysis, we are now witnessing the emergence of a parallel phenomenon on
the production side: distant writing.^2
Distant writing refers to a literary creation practice wherein human authors
function primarily as narrative designers, while LLMs based on Large Language
Models^3 (LLMs) perform the actual writing.^4 Unlike traditional authorship, distant
writing—also occasionally termed wrAIting^5 in this article to distinguish it from close
writing (the ordinary kind)—positions the author not as the direct textual producer, but
as the architect of narrative possibilities, responsible for specifying requirements,
affordances, and constraints, and curating the LLM-generated content. This shift
represents not merely a technological intervention in the writing process but a
fundamentally different conceptualisation of what it means to author a text.^6
In this article, I examine distant writing as an emerging literary practice,
discussing its theory, methodology, and implications for authorship, creativity, and

(^1) For an introduction and a bibliography see the very valuable work by Scott Newstok here
https://www.closereadingarchive.org/ 2
3 This article could be read as a long commentary to, and discussion of (Flusser 2011).^
Strictly speaking LLMs are the core models (e.g. a version of GPT) that support the user-facing
applications (e.g. a version of ChatGPT) built on top of them. In the article, this distinction is
unnecessary, and I do not use it. Instead, I use LLMs to mean both the models and the applications
built on top of them. 4
I am not the first, by far, to call attention to a phenomenon by now widespread, for earlier references
see {Rodley, 2014 #33} and more recently (Bajohr 2024). Many thanks to Scott Newstok for calling my
attention to the work by Leonardo Flores or Alessandro Giammei. The expression “distant writing” is
also common and has been used before, sometimes in ways unrelated – e.g. distant writing as something
that happens with the telegraph (Roberts 2015)or long distance correspondence – sometimes in ways
that are relevant but pre-LLM or different from how it is used in this article, see (Flores 2020), (Fortunati
and O’Sullivan 2021) 5.
Emmie Hine is right: this neologism is awful. So, I use it sparingly, just to avoid confusion, and I hope
it will not catch. 6
For a fascinating overview of “computational writing” from its early stage to some of the topics I
discuss in this article and the analysis of different kinds of “distances” see (Bajohr 2024).

narrative structure. Drawing on recent experiments (Floridi 2025b), I argue that distant
writing expands rather than replaces human creativity within a design framework,
prompting essential questions about the boundaries and future of literature in an AI-
driven culture (Floridi 2024, 2025a).

2. From Distant Reading to Distant Writing: Theoretical Frameworks
Moretti’s distant reading emerged as a response to the limitations of close reading
when faced with the vastness of world literature. As he argued
distant reading ... allows you to focus on units that are much smaller or much
larger than the text: devices, themes, tropes—or genres and systems (Moretti,
2000, p. 57).
Through computational analysis, patterns emerge across texts that would otherwise
remain invisible to traditional methods of literary analysis. Big data, in essence, reveals
micro-patterns that demand specialised tools for their identification and analysis
(Floridi 2012a). Distant writing operates on analogous principles but inverts the
direction of the computational engagement. Rather than analysing existing texts to
identify existing patterns, distant writing designs new patterns—through requirements,
affordances, constraints, and other parameters—to instruct computational systems to
generate new texts from existing ones. Where distant reading employs computation to
expand interpretative options, distant writing uses it to open creative opportunities.
Given this inverse relationship, it is important to clarify a key point here that
will become more evident later in the article. Moretti accurately defines distant reading
in terms of feasibility relative to close reading, characterising it as an interpretative
method best suited to tasks that close reading either cannot accomplish or would
significantly struggle to achieve. This does not mean distant reading cannot replace
close reading, but doing so forfeits its unique advantages. It is comparable to using
virtual reality for a concert one could attend in person—a lesser experience. Virtual
reality is most valuable when enabling otherwise impossible experiences, not when
merely replicating achievable ones (Floridi 2022). Likewise, distant writing enables
narratives that would otherwise be impossible or difficult to achieve. This does not
mean that an author cannot fruitfully use distant writing to LLM-generate texts that
could be produced by close writing, but that this practice would fail to capture what is

unique in the process adopted. Like distant reading, distant writing is a matter of
expanded feasibility, not just increased efficiency or effectiveness, although the latter
features are certainly a crucial part of the success of the LLM-based production of
texts.
The theoretical framework for distant writing draws upon several established
domains. Let me anticipate here what I shall discuss in more detail below. From modal
logic, it adopts the conception of narrative as a boundless space of possible worlds
(Ryan 2006) constrained by coherence rather than mere consistency (Ryan 1991), and
narrative structuring principles extensively discussed in narrative theory (Genette 1980,
Herman 2002, Ricoeur 1984-1988). From design theory, it borrows the understanding
of creativity as a process of exploration within affordances and constraints (Gero
1990). And from computational creativity, it takes the notion of the human-computer
partnership as generative of possibilities inaccessible to either in isolation (Boden
2004). However, distant writing also introduces novel theoretical considerations, like
what one might term the “isotropy of the narrative space”: the principle that any
narrative domain is equally workable in any direction, provided coherence is
maintained. Unlike traditional conceptions of narrative as having preferred directions
or natural progressions, distant writing reveals a boundless number of alternative paths
in a constructivist sense. This does not mean the paths are already there and only need
to emerge. Following a classic distinction in the philosophy of mathematics between
boundless and infinite ,^7 distant writing, as I present it here, accepts boundless (potentially
infinite) paths but rejects the existence of actual (completed) infinite paths. There are
as many narrative patterns as needed, and more can always be added if our operations
require more, but there is no infinite set out there, to be discovered independently of

(^7) In the philosophy of mathematics, particularly from the constructivist viewpoint, the terms
“infinite” and “boundless” have distinct meanings. “Infinite” typically refers to a completed
totality, an actual infinity—something that exists fully realised as a finished entity, e.g., classical
mathematics’ set of all natural numbers. “Boundless”, by contrast, denotes a potential
infinity—an indefinite process or construction that can always continue further, but is never
actually completed or totalised, e.g., the constructivist conception of the natural numbers,
where one can always produce the “next” number, but the entire infinite set never exists as a
completed whole.

our “writing”. To use one more distinction from the philosophy of mathematics,
distant writing is not Platonist, it is Constructivist:^8 all stories are created, they are not
mind-independent realities to be discovered or revealed. The poet may disagree, here
is what Jan Skácel wrote:
Poets don’t invent poems
The poem is somewhere behind
It’s been there for a long long time
The poet merely discovers it.
Jan Skácel was one of the best poets writing in Czech. Milan Kundera cites him and
discusses these lines in his essay “Somewhere Behind”.^9 There we can also read that
“novelists draw up the map of existence (italics in the original) by discovering this or that
human possibility” (see Part Two). But this is a twofold confusion. Because the writer
designs , which is a third category between inventing in one’s own mind and discovering in
the outside world. Design is the essential creative category that characterises our time,
as the ability to make the inside and the outside relate, like a key and a locker. And
because the writer mistakes, like a mathematician,^10 the strength of logic (necessity) for
the power of ontology (inevitability). As Hemingway puts it
The laws of prose writing are as immutable as those of flight, of mathematics,
or physics. (Hemingway 1984), p. 77.
Yet the narrative constrains (resistance) and affordances (facilitation) perceived are not
of some-thing out there, but of a some-how that forcefully links the relata. Vilém Flusser,
one of the greatest Czech philosophers of his generation, understood this better and
earlier than most. Let me quote him at length (the original text, in German, was
published in 1987):

(^8) I must acknowledge here the influence of Michael Dummett’s version of mathematical
intuitionism on my way of thinking about narratives, although we never spoke about
narratives, only about logical proofs.
(^9) Jan Skácel was one of the best poets writing in Czech. I came to know about him thanks to Milan
Kundera, who cites and discussed these lines in “Somewhere Behind”, Part Five of (Kundera 1988).
Much as I admire him and this book in particular, I disagree that “novelists draw up the map of existence
by discovering this or that human possibility 10 ”.
The analogy is also Kundera’s, who in Part Four write, discussing his understanding of the structural
properties of his work The Joke , that “It was a revelation to me. In other words, that ‘mathematical
system’ emerges completely naturally as a form of necessity, with no need for any calculation”.

His attitude [the poet’s] to a poem is no longer that of the inspired and intuitive
poet but that of an information designer [my italics]. He relies on theories and no
longer works empirically. Such an informatic approach to poetry has long been
in preparation. [...] All our conceptions of poets favored by the muse must
yield to a conception of the poet as a language technician. Poetry will be
desanctified. [...] The new poet, sitting at his terminal and waiting expectantly
to see which unanticipated word and sentence formation appear on the screen,
is gripped by a creative delirium no less intense than the one a writing poet felt
in his struggle with language. Each time a technical threshold is crossed,
observers have the sense that technology is getting the upper hand, and each
time, it turns out that the new technology opens new creative possibilities. [...]
On one hand, there will be artificial intelligences that speak, presenting a
continuous program of new poems in keeping with their programs. And on
the other hand, information designers [my italics] will, with the help of a
permutation game, cause poems – coded alphabetically or not – to light up on
our screens before us at a breathtaking pace, like some kind of artificial Elliot
or Rilke. (Flusser 2011), pp. 75 - 76
Figure 1 summarises the distinctions just introduced that frame the following sections.

Figure 1 : Close and distant reading and writing

Mode of Engagement
Reading Writing
Gr
an
ula
rit
y
Close
Close Reading
Human interpreter engages directly
with individual texts.
Focus on nuances, e.g., language,
rhetoric, character psychology.
Exemplified by line-by-line exegesis
in traditional literary criticism.
Close Writing
Human author composes text directly.
Emphasis on individual style, voice,
and manual revision.
Traditional authorship process:
drafting, editing, and polishing by the
writer.
Distant
Distant Reading
Computational analysis of large
corpora to detect macro-patterns,
e.g., themes, tropes, genre
systems.
“Big data” methods reveal patterns
imperceptible to close reading.
Distant Writing
Human as the designer of narrative
constraints; sets requirements through
prompts; curates AI output.
AI (LLMs) as the executor of text.
Enables narratives infeasible by close
writing alone.
Inverted Relation
3. Methodology of Distant Writing: Designing through Prompting
The methodology of distant writing centres on designing the textual input that
generates the desired textual output. This approach reconceives writing as a systematic
interrogation: “wrAIting” becomes the art of crafting targeted prompts that guide an
LLM toward generating the desired narrative output. The precision of these
prompts—the correctness and accuracy of the Socratic maieutic exercised by the
authoritative designer to nudge and ultimately force the LLM to generate the desired
output—determines the quality of the narrative. Poor prompts yield incoherent texts,
while well-crafted prompts may produce sophisticated, thematically consistent literary
outcomes. The iterative process, known as progressive refinement ,^11 leads to the outcome
ultimately released by the author, whose approval bears all the intellectual responsibility
for the shared content. The methodological process is well known to anyone who has
used any LLM, so I shall only briefly summarise it. It typically involves seven stages,
although each writer will have their preferred approaches and solutions:

Conception and Development: this preliminary stage encompasses all activities
involved in conceiving and developing the narrative idea. Although interactions
with one or more LLMs may offer heuristic assistance here, they remain external
to the core distant writing process; at this point, the author simply determines
whether adopting distant writing is the most suitable approach to realise the
intended narrative.
Requirements Formulation: establishing the affordances, constraints, and other
parameters to guide the narrative generation, including style, themes, characters,
and plot elements.
Prompt Engineering: formulating precise requirements as effective prompts
designed to elicit specific responses from the LLM. Prompts can be crafted directly
using natural language (NL) or, increasingly, by employing specialised systems that
convert NL prompts into formats optimised (markup) for use by LLMs.
(^11) Progressive refinement is a methodology in computer science and software development
where a solution is gradually improved through iterative steps, starting with a basic version
and incrementally enhancing it until reaching the desired quality or functionality.

WrAIting: LLM-based generation of text.
Progressive Refinement: in this iterative process, the author reviews LLM-
generated text, identifies shortcomings (often with the help of LLMs, see next
step), and refines prompts or formulates new ones to address the relevant issues,
re-wrAIting the text. The process does not end automatically since further
refinements are always possible. Instead, it is the author’s responsibility to
terminate it.
Validation and Verification: using an LLM, possibly a different one, to ensure the
generated narrative maintains coherence, plausibility, and adherence to established
requirements. Currently, this is more an art than a science, as there are no
metrics/benchmarks for this. For example, and at the time of writing, some LLMs,
like the Claude series, seem better at writing narratives, while others, like the GPT
series, seem better at validating and verifying them, but they still need to be
prompted correctly. Hallucinations are always possible, so this is a stage where close
reading and a lot of critical thinking become crucial.
Curation and Assembly: selecting, arranging, and potentially modifying generated
content to form a cohesive final work.
The process reveals an important distinction between traditional writing and distant
writing: the latter operates through a logic of requirements rather than a logic of
composition. This is not surprising: the logic of design is the logic of requirements
(Floridi 2017b), and in distant writing, the answer required follows the requiring
question: the better the latter is, the more the answer is constrained and closer to the
original design. To quote Flusser again:
All writing orderly, and that leads directly to the contemporary crisis in writing.
For there is something mechanical about the ordering, the rows, and machines
do this better than people do. One can leave writing, this ordering of signs, to
machines I do not mean the sort of machines we already know, for they still
require a human being who, by pressing keys arranged on a keyboard, orders
textual signs into lines according to rules. I mean grammar machines (my italics),
artificial intelligences that take care of this order on their own. Such machines
fundamentally perform not only a grammatical but also a thinking function,
and as we consider the future of writing and of thinking as such, this might
well give us pause for thought. (Flusser 2011), p. 6
4. Authorship Reconfigured: The Meta-Author as Responsible Designer
In conventional literary production, the author functions as both the conceptual
architect and the textual executor: the person who both conceives and writes the work.
In distant writing, these functions are decoupled, with the human operating as what
might be termed a “meta-author” who designs but does not directly produce the text.
Historically, this decoupling has happened before. We find it natural and
obvious in architecture, fashion, industrial design, music, and all cases in which the
authors are not (although sometimes they can be) the executors of the projects they
design. Even in the history of art, great artists like Raphael would sometimes only
design or sketch a painting, while the assistants in his workshop would execute it, with
perhaps a final touch by Raphael himself. Nowadays, the decoupling is magnified in
scope and depth as the effect of a broader trend characterising digital innovation’s
cleaving power (Floridi 2017a): the ability of digital technologies to aggregate or
disaggregate phenomena that were considered in the past entirely independent or even
unrelated—e.g., personal data and personal identity, when talking in terms of data
subjects or individual profiles—or so strictly connected to be indivisible—e.g. law and
its territoriality.
Culturally, the decoupling between text designer and text executor raises
challenging questions about authorship, creative agency, attribution, and ownership.
For example, to whom should we attribute a work of distant writing? The human
designer who established the requirements? The LLM that generated the text? Or
perhaps the original authors whose works might have been used to train the model?
The concept of a meta-author as a designer of textual content acknowledges
the human’s creative input while recognising the distributed nature of agency in distant
writing. It suggests a model of authorship that is neither entirely human nor entirely
artificial but emerges from the interaction between human intention and agentic
capability (Floridi 2025a). I just mentioned that this reconfiguration of authorship
parallels developments in other creative fields. The crucial notion here is that of
responsibility , not ownership. It is the designer that is accountable for the final text. The

comparison to architecture seems particularly apt, as architects have long been
recognised as the creative forces behind buildings they never physically constructed
but were responsible for. Bernini neither made nor put in place the 284 columns in St
Peter’s Square, yet it is unquestionably Bernini’s Colonnade. Distant writing extends
this model to literature, positioning the meta-author as a designer whose creative vision
is executed through technological means that shift the boundaries of the feasible.
Hemingway might have agreed, even if criticising my Baroque choice of Bernini:
prose is architecture, not interior decoration, and the Baroque is over
(Hemingway 1984), p. 72.
There is a final difference worth stressing here: Raphael and Bernini may not have
delivered some of their masterpieces had they not been commissioned by Pope Julius
II or Pope Alexander VII. On the contrary, digital technology, LLMs included, further
empowers individuals to be their own masters. Meta-authors can be their own
commissioners.
All this means that the so-called “death of the author” as a writer (Barthes
1967, Foucault 1969)—even if exaggerated and correctly questionable in its radical
interpretations (Hirsch 1967, Fish 1980, Burke 2010, Farrell 2017)—may be followed
by the birth of the author as a designer and co-producer of a text, who initiates,
supervises and is responsible for the collaborative wrAIting with LLMs, past generations
of writers on whose works the LLMs are trained, past as well as potential readers, and
at some point past LLMs as well.

5. The Modal Space of Narrative: Boundless Isotropy and Experimentation
One of the interesting contributions of distant writing is its reconceptualisation of the
narrative space as boundless. While narrative possibilities are inexhaustible, they do
not exist until they are realised, and to be realised, they need to comply with
requirements of coherence, plausibility, and other parameters established by the meta-
author. Drawing on modal logic, distant writing shows more clearly what was always
obvious, at least implicitly: consistency , that is, the absence of contradiction, is necessary
but insufficient. Coherence is also required, understood as the consistent co-presence
of meaningful connections. For example, consistency is necessary but insufficient for

a meeting between Hamlet ( Hamlet ) and Caliban ( The Tempest ), even if an LLM can
easily produce the scenario and dialogue in
Table 1.

Hamlet, the melancholic prince of Denmark, is shipwrecked on Prospero’s
enchanted island while fleeing Denmark. Wandering the island, he
encounters Caliban, Prospero’s resentful servant, who mistakes Hamlet for
another tormenting spirit.
Caliban: What cruel spectre art thou, come to plague me further?
Hamlet: I am but a man, though burdened by the ghost of my father’s will. And what art
thou, this wretched creature, half man, half beast?
Caliban: I am the rightful lord of this isle, usurped by a wizard most vile. But thou, pale
stranger, art no lord, just a shadow lost as I am.
Hamlet: A shadow indeed, for my purpose is as fleeting as the wind. Tell me, monster,
dost thou suffer as I do, trapped between action and despair?
Caliban: Suffer? Aye, each moment, tormented by the lash of men who call themselves my
betters. I curse, I rage, but what fruit bears my wrath?
Hamlet: Then we are kin, thou and I, for I curse my fate yet lack the will to end it.
Table 1 Example of LLM-generated Shakespearean dialogue
A narrative must also be plausible regarding features like characters’ ages, times and
places of the events, social classes, distances travelled, or financial means. If an LLM
is not carefully instructed to curate it, it will fail to implement this, sometimes very
subtly.
I already mentioned that this understanding of narrative as a modal space aligns
with possible worlds theory in literary studies (Pavel 1986, Ryan 1991), but distant
writing adds a crucial insight: the isotropy of narrative space. Traditional narrative
theories often imply preferred directions or inherent progressions within narrative
structures. By contrast, distant writing shows that narrative space can be shaped in any
direction desired, similar to how marble can be sculpted freely, provided coherence is

maintained. This isotropy has profound implications for literary creation. It suggests
that any story is only one possible realisation of a boundless space of narrative
possibilities, although not all narratives are genuinely possible. Other patterns can be
designed within the same space, and distant writing provides access to these alternative
realisations in ways traditional writing cannot, or struggles to do. This introduces a
final remark about experimentation. Distant writing can be a powerful tool to navigate
possible spaces of narrative development, in at least two ways.
Distant writing can test counterfactuals: whether the narrative logic of an
existing narrative is robust. For example, one can use it to test whether the narrative
logic of Pride and Prejudice holds under perturbation—i.e., to check whether altering
constraints, such as character, chronology, or causality, break the internal coherence
and consistency of the story. For example, in a variation of Pride and Prejudice , what
would follow if Elizabeth accepted Mr. Collins’s proposal? This process uses the LLM
not to wrAIte Austen, but to stress-test her text by simulating variant developments.
The LLM helps explore narrative counterfactuals that would be impossible or too
laborious to model manually, showing where Austen’s plot is flexible or intricately
rigid. Ideally
the stories are written so tight and so hard that the alteration of a word can
throw an entire story out of key. (Hemingway 1984), p. 79
However, distant writing can also help explore alternative or unimplemented narrative
paths—that is, hypothetical scenarios that are allowed and plausible but unrealised
developments—within the existing constraints of a text. Here are four examples
(including a counterfactual one) using Emma :

Counterfactual: “If Frank Churchill had not been secretly engaged to Jane
Fairfax, Emma would have fallen in love with him.” This scenario contradicts
the story’s actual events, imagining an alternative outcome where Frank’s
sincerity leads to a romantic relationship with Emma, altering her path to Mr.
Knightley.
Semifactual: “Even if Mr Elton had not proposed to Emma, Harriet would
still have been heartbroken.” This scenario keeps the actual outcome (Harriet’s
heartbreak) constant but imagines a change in the cause, suggesting her
heartbreak would persist due to her insecurity or misplaced affections.
Forward-Looking Indicative Hypothetical: “If Mr Knightley were to express
his feelings to Emma tomorrow, she would likely realise her own love for him.”
This scenario speculates on a plausible future event based on the story’s actual
development, without contradicting or altering established facts.
Open Hypothetical: “What if Emma lived in a modern world where
matchmaking was done through dating apps?” This scenario reimagines the
story in a speculative context, disconnected from the constraints of the original
narrative, exploring how Emma’s meddling might translate to a technological
setting.
As James Wood writes when discussing Aristotle’s “original formulation of mimesis”:
Hypothetical plausibility – probability – is the important and neglected idea,
here; probability involves the defence of the credible imagination against the
incredible. This is surely why Aristotle writes that a convincing impossibility in
mimesis is always preferable to an unconvincing possibility [...] Internal
consistency and plausibility (what I have called above “coherence”) then
become more important than referential rectitude. (Wood 2008), p. 179.
The point is not what “hypothetical plausibility” can be tested, but why it would be
interesting to test that rather than another; that is an author’s choice and decision.
Writing, and more visibly wrAIting, is ultimate a matter of taking and allocating
author’s responsibility.
6. Minor Characters and Network Narratives: The Connected Nature of Fiction
Encounters (Floridi 2025b) provides an application of distant writing as the exploration
of minor characters from existing literature. By extracting secondary figures from
canonical works and placing them coherently in new narrative contexts, distant writing
reveals the connected nature of fiction as a single possible directed graph^12 where any
character can be connected to any other character through a finite number of steps
(other encounters). As Frigyes Karinthy famously suggested in Chains (Karinthy 1929),
humans are all connected through just a few acquaintances. The degree of separation
(^12) In Encounters , the directed graph is a simple directed path, but nothing prevents more
complex structures.

is likely to be higher than 6 for many characters in the narrative space, but they, too,
are all connected; one only needs to design the right meetings.
This level of abstraction highlights minor characters as nomina infinita , defined
by what they are not. Unlike essential major characters, minor figures represent open
possibilities, connecting narratives across literary worlds. By focusing on these
underdetermined nodes in the narrative network, distant writing can create
connections between seemingly disparate literary worlds, revealing that in the narrative
space, no single story is entirely disconnected from any other story. Some
neighbourhoods of the narrative space are more densely populated than others and
cab be more easily exploited. Thus, in Encounters , I experimented with minor characters
from English-written novels to deal with an easier case. But more adventurous
explorers have the languages and the stories of all humanity at their disposal. Because
this connected nature of fiction allows distant writing to create narratives that span
across different authors, periods, genres, regions and cultures, weaving together a
tapestry of literary references and relationships that can be truly universal.
These network narratives represent a form of literary exploration previously
difficult or impossible to achieve. Traditional comparative literature might identify
influences and parallels between works, but distant writing actively creates new
pathways between texts, demonstrating the interconnectedness of the literary
landscape through generative rather than analytical means. The positive effect is that
distant writing expands the possibility of intertextual reappropriation, the reuse of
characters from other works, which authors have exploited to create new and
innovative stories about specific, familiar characters, with fresh perspectives or
expanded narratives.^13 Humanity’s inexhaustible semantic capital (Floridi 2018) is not only

(^13) Famous examples of intertextual reappropriation include: James Joyce reusing Stephen
Dedalus, the protagonist of A Portrait of the Artist as a Young Man , as one of the central
characters in Ulysses ; Jean Rhys reimagining Bertha Mason, the “madwoman in the attic” from
Charlotte Brontë’s Jane Eyre , in her novel Wide Sargasso Sea , telling her tragic backstory and
humanising her character; Gregory Maguire giving a new perspective to the Wicked Witch of
the West from L. Frank Baum’s The Wizard of Oz in Wicked , portraying her as a misunderstood
and complex character. And Tom Stoppard focusing on Rosencrantz and Guildenstern, two
minor characters from Shakespeare’s Hamlet , in Rosencrantz and Guildenstern Are Dead ,
presenting the events of Hamlet from their point of view.

unique but also undivided and non-hierarchical, and distant writing can help put it to
good use, for creative and hermeneutical purposes, although not unproblematically, as
I shall indicate presently.

7. Dataprint: The Stylistic Signature of LLMs
A significant observation from experiments in distant writing is the emergence of what
might be termed an “LLM stylistic signature”: identifiable patterns, rhythms, and
structures that characterise texts generated by specific models regardless of the stylistic
parameters established in prompts. Regardless of quality, and independently of the
specific stylistic choices (e.g., emulating Austen) determined by the human designer,
LLMs exhibit common patterns—such as rhythm, narrative structure, distinct ways of
initiating, developing, and concluding a story, lexical preferences^14 —forming an
identifiable literary pattern that can be termed their dataprint.^15
Clearly, the term “voice”, often used to describe human authors, would
constitute an inappropriate anthropomorphism (Floridi and Nobre 2024), but it is
close in meaning, as it refers to the distinctive and consistent style, tone, and
perspective expressed throughout a writer’s works, making their writing recognisably
unique. The difference is that “voice” encompasses not only implicit features of which
an author may be unaware—for example, recurring thematic interests, underlying
beliefs, attitudes, and emotional stances towards the topics they address—but, above

(^14) A classic is the overuse of the verb “to delve” (Juzek and Ward 2024).
(^15) Dataprint is not a neologism, but a technical term in computer science. I am borrowing here
to refer to the stylistic signature of an LLM and hence to stylistic traces that can be identified
(often through other LLMs) in the output of a distant writing process. This is not far from the
original meaning, according to which a dataprint (also known as a data fingerprint or digital
fingerprint) is a unique, structured representation derived from a dataset or digital content,
typically generated through computational methods such as cryptographic hashing, pattern
extraction, or feature encoding. In computer science, dataprint is designed to capture key
identifying characteristics or patterns of the data, enabling verification, authentication,
comparison, traceability, and detection of alterations without requiring full exposure or
reproduction of the original dataset. So dataprints are commonly used in contexts involving
data provenance, integrity verification, cybersecurity, plagiarism detection, digital forensics,
and machine learning to uniquely characterise data instances, facilitate deduplication, or
support anomaly detection. In distant writing, all this remains true, but the emphasis is on
style.

all, the deliberate choices an author makes concerning vocabulary, sentence structure,
rhythm, or punctuation, which collectively establish their narrative identity. A “voice”
manifests through the specific rhetorical strategies, such as figurative language and
imagery, that an author habitually employs, contributing significantly to the
individuality and coherence of their literary output. Once more, distant writing
decouples these two aspects, the intended and unintended sides of an author’s voice.
LLMs have no intelligence or mental states, hence no deliberate choices, but they still
show unique “wrAIting styles” that transcend the specific stylistic constraints imposed
by prompts. Their dataprints do not show personal intent, emotions, or experiences
to influence. They are regular, emergent features, often invariant and hence
informative, that can identify them as the source of a text and distinguish them from
each other. The reader who may be somewhat sceptical may wish to consider that it is
because of their dataprints that we can guess, with a high level of probability, whether
an LLM has written a text, and how a simple experiment with two LLMs will show
that they differ remarkably in style when responding to the same storytelling prompt.
AI’s dataprint allows distant reading and computational approaches—
stylometric analysis, lexical frequency analysis, syntactic complexity measures, and
supervised classifiers trained specifically on outputs from different models—to detect
and attribute LLM-generated authorship. For example, when not prompted for a
specific style, currently the GPT-series dataprint tends to be more lyrical, expansive, and
introspective, favouring a nuanced emotional register. It crafts narratives rich with
sensory details and elaborate, evocative descriptions, often weaving introspection
seamlessly into the storytelling. The prose typically exhibits more complex syntactic
patterns and a refined, sophisticated lexicon, underscored by a clear psychological
depth. This dataprint prioritises emotional resonance, character-driven plots, and
structured narrative arcs designed for emotional gratification. The Claude-series
dataprint, by contrast, tends to be more direct, lean, and pragmatic, characterised by
economical language and narrative efficiency. It typically employs a minimalist prose,
emphasising concise action and forward momentum rather than intricate descriptions
or introspective digressions. The storytelling is straightforward, often letting
characters’ actions and minimal dialogue imply emotional states, creating immediacy
and clarity without extraneous ornamentation.

So how does a dataprint emerge? Since an LLM is not inherently creative or
expressive in the same way as a human, its stylistic signature emerges from a
combination of its design, training data, and the way it generates responses. Here are
the main factors behind a specific stylistic signature:

Training Data Influence: the corpus of text used to train the LLM shapes its
tone, style, and depth of knowledge. For example, an LLM trained on formal
academic texts may exhibit a precise, professional tone, while an LLM trained
on conversational or creative texts might produce more casual or imaginative
responses.
Output Patterns: LLMs often develop distinct response tendencies based on
their architecture and training. These patterns might include vocabulary
choices (e.g., formal vs. colloquial language), sentence structure (more or less
concise or elaborate sentences), tone consistency (e.g., neutrality, enthusiasm,
or empathy), formatting style (e.g., bullet points, numbered lists, or essay-style
responses).
Design and Parameters: these are the design choices of the LLM’s developers,
such as its model size, fine-tuning, and use cases, that impact its stylistic
signature. For example, the GPT-series often aims for clarity and coherence,
making their responses accessible and user-friendly. Models specialised for
creative writing might exhibit a more imaginative or poetic style.
Limitations and Repetitions: repeated stylistic quirks or limitations can also
form part of an LLM’s signature, such as a tendency to hedge statements (e.g.,
“it depends...” or “there are many factors...”), recurrent sentence starters, such
as “in summary...” or “for example...”, or an over-reliance on specific
structures, like defining concepts before providing details.
Fine-Tuning and Applications: fine-tuning an LLM for specific industries or
tasks can add unique stylistic elements. For example, a legal-focused model
may consistently use formal and legalistic language, while a model fine-tuned
for creative writing might emphasise metaphor, imagery, and narrative flow.
The emergence of LLM stylistic signatures raises intriguing questions about the nature
of style itself. Traditional literary theory has often conceived of style as expressing an
author’s unique sensibility or worldview (Barthes 1967). But if an LLM can develop a
recognisable style without consciousness or intention, perhaps style is better
understood as a set of statistical patterns and regularities, some of which are intended
and purposefully sought (“voice” in a strict sense), while others are unintended and
only emergent (dataprint). Humans have both, but an LLM has no voice, only a
dataprint.
The implications extend beyond theoretical interest. If LLMs develop
increasingly distinctive styles, future literary historians might analyse these
computational signatures just as they currently study the styles of human authors. We
may eventually speak of “early Claude” versus “late Claude” periods, or identify
influences between different LLMs in the same way we trace influences between
human writers.

8. Creating New Literary Genres: LLMs and the Evolution of Narrative Forms
Distant writing not only reconfigures existing literary forms but also creates the
conditions for new genres to emerge. Technological changes have often facilitated
genre evolution throughout literary history, from the epistolary novel enabled by postal
systems to the stream-of-consciousness technique influenced by new psychological
theories. Distant writing represents a similar technological catalyst with the potential
to generate novel literary forms that would be difficult or impossible to achieve
through traditional human writing.
One emerging genre might be “multiverse literature” (Booth 2024), texts that
explore multiple narrative paths simultaneously rather than sequentially. While
hypertext fiction experimented with reader-directed branching narratives, distant
writing enables the efficient creation of vast narrative landscapes where countless
variations of a story coexist. With distant writing, authors could design works that
purposefully explore multiple carvings of the same narrative marble, presenting readers
with a cohesive multiverse rather than a single storyline.
Another potential genre is what we might call “network narratives”—texts that
systematically explore the interconnections between seemingly disparate fictional
worlds. The Encounters experiment, bringing together minor characters from different
works, represents an early example of this approach. Such narratives could evolve into

complex literary ecosystems that map the “connected network” of fiction, creating
meta-stories that exist in the spaces between established texts.
Distant writing also enables new forms of collaborative authorship that
transcend traditional co-writing. “Conversational literature” might emerge as a genre
where human designers and LLMs—which could also impersonate dead authors,
imagine an interview with Italo Calvino on “mechanical writing”^16 —engage in
structured dialogues that generate narratives through iterative exchanges. Unlike
traditional collaborative writing, which typically involves humans working sequentially
or in parallel, conversational literature would represent a genuine fusion of human and
machine creativity, a dialogue between designer and system that produces texts neither
could create independently.
The concept of originality itself requires reconsideration in this context. The
Romantic notion of originality as ex nihilo creation has long been challenged by
postmodern theories of intertextuality, which recognise all texts as “mosaics of
quotations” (Kristeva 1980, Hutcheon 1988, Eco 1989). Distant writing makes this
intertextuality more explicit and systematic. Besides, attempting to create something
entirely new, distant writing can also embrace “deep remixability” (Manovich 2007),
simultaneously recombining cultural elements at multiple levels. We saw that this
recombinatorial approach to originality shifts emphasis from creating unprecedented
content to discovering unprecedented connections. As (Boden 2004) distinguishes
between P-creativity (psychological novelty, new to the individual) and H-creativity
(historical novelty, new to human history), distant writing introduces what one might
call C-creativity : combinatorial novelty that emerges from the systematic exploration of
possible connections within the space of narrative coherence.

(^16) Italo Calvino (1923–1985) was known for his experimental narratives and imaginative
literary style. In his essay Cybernetics and Ghosts (1967, collected in (Calvino 1986) he speculates
on literature as a combinatorial and computational process, suggesting that machines might
one day generate literary works. Calvino’s reflections anticipated contemporary discussions
about computational creativity, LLM-generated literature, and mechanical writing. Calvino was
an active member of the Oulipo (Ouvroir de Littérature Potentielle), a group of writers and
mathematicians who created works using constrained writing techniques. See also (Calvino
1987), (Usher 1995).

These emerging genres challenge traditional literary categories, like content,
form, or effect, by focusing instead on their compositional logic or the human-AI
collaborative process. This would represent a fundamental shift in classifying and
understanding literary forms, moving from product to process-centred taxonomies.
Such a shift would align with an ontology that values relational, network-oriented
thinking instead of substantial, mechanical-oriented one (Floridi 2008, 2012b). I
defined it elsewhere as post-Vitruvian (Floridi 2024).
As these new genres evolve, they may transform readers’ expectations and
competencies. Just as modernist literature required readers to develop new interpretive
strategies, LLM-generated genres might foster new kinds of literary literacy—abilities
to recognise patterns across narrative multiverses, trace connections in network
narratives, or appreciate the specific qualities of human-LLM collaborative creation.
The originality of distant writing may ultimately lie not only in the texts it produces
but also in the new modes of experimental reading it enables.

9. Multiverse Literature: Exploring Narrative Branching at Scale
The concept of multiverse literature (Booth 2024) deserves further exploration as a
genre enabled by distant writing. While traditional narratives typically follow a single
path through a space of possibilities, multiverse literature systematically explores
multiple paths simultaneously, presenting not a single story but a coherent system of
related stories that branch and converge in complex patterns. Traditional attempts at
branching narratives—such as choose-your-own-adventure books, hypertext fiction,
or interactive games—have been limited by practical constraints. The labour involved
in manually creating multiple narrative branches quickly becomes prohibitive as the
number of decision points increases, leading to what game designers call the
“combinatorial explosion problem.” Even sophisticated interactive fiction must
severely limit the number and significance of branching points to remain manageable.
Distant writing fundamentally changes this equation. LLMs can efficiently generate
multiple narrative continuations from any given point, enabling the exploration of
narrative possibility spaces at previously impossible scales. Multiverse literature
exploits the isotropy of narrative space, intentionally designing multiple narrative paths

simultaneously and regarding each path as equally valid, rather than privileging a single,
linear progression.
The structure of multiverse literature might take several forms. The simplest
would be a “parallel narratives” approach, where a single premise generates multiple
distinct storylines that never converge. More complex would be a “branching and
converging” structure, where narratives diverge at key decision points but may later
reconverge, creating a network of interconnected possibilities rather than a simple tree.
Even more ambitious would be full “quantum narratives” where multiple states exist
simultaneously in superposition, with certain elements remaining constant across
versions while others vary.
What distinguishes LLM-generate multiverse literature from earlier
experiments in branching narratives is not just scale but coherence. Rather than
presenting disconnected alternative paths, multiverse literature would emphasise the
relationships between variants, treating the system of possibilities as a unified work
rather than a collection of separate stories. This systemic approach might reveal
patterns and structures invisible in any single narrative instantiation, creating a meta-
level of meaning that emerges from comparing alternatives.
The reading experience for multiverse literature would differ significantly from
traditional linear reading. Readers might navigate the narrative space through various
interfaces—perhaps selecting specific branches to explore, perhaps experiencing
multiple versions in parallel, or perhaps following algorithmically generated paths
based on their preferences. Like in old hypertextual experiments, the boundaries
between author, reader, and text become more fluid, as readers participate in
determining which narrative possibilities are actualised in their experience.
This form of literature connects to philosophical questions about contingency,
necessity, and possibility. As modal logic distinguishes between actual, possible, and
necessary, multiverse literature makes these distinctions explicit within a narrative.
Some elements might appear in all versions (necessary), others in only some versions
(contingent), while still others remaining merely possible, creating a hierarchy of
narrative inevitability that traditional storytelling can only implicitly suggest. The
implications for characterisation are significant. Characters in LLM-generated
multiverse literature would exist not as fixed entities but as possibility spaces—

collections of potential traits, decisions, and developments that manifest differently
across variants. This approach aligns with the previous discussion of minor characters
as nomina infinita , defined more by their possibilities than their actualities. In multiverse
literature, all characters retain something of this infinitude, never fully determined by
any single narrative path.
As a critical framework, multiverse literature might draw on quantum physics
(Ryan 2006, Booth 2024), modal logic, and possible worlds theory to develop new
vocabularies for discussing narrative. Critics might analyse not just what happens in a
story but the probability distribution of possible happenings, the phase space of
character development, or the strange attractors that pull disparate narrative branches
toward common outcomes despite differing paths.
While early experiments in multiverse literature might focus on explicitly
branching narratives, more sophisticated works could develop subtler approaches to
multiplicity—perhaps embedding alternative possibilities within seemingly linear texts,
creating palimpsest-like layers of potential meaning, or developing formal techniques
for suggesting the ghost-presence of untaken paths within the chosen one. The full
artistic potential of this genre remains to be explored as distant writing techniques
continue to evolve.

10. Pedagogical Implications: Teaching Creative Writing in the Age of LLMs
The emergence of distant writing invites reconsidering how creative writing is taught.
Traditional creative writing pedagogy has focused primarily on developing skills in
direct textual production: crafting sentences, structuring narratives, developing
characters, and refining style through practice and revision. Distant writing, however,
shifts the emphasis from execution to design, requiring different skills and approaches.
In this new paradigm, creative writing education might evolve in several directions.
Programs may need to incorporate “prompt engineering” as a core
competency, part of a broader shift towards multiliteracies and digital literacy skills
essential in contemporary education (Selber 2004, Gee 2007). Since distant science and

distant scholarship follow directly from non-fictional distant writing,^17 it is already
essential to teach students how to formulate effective requirements that guide LLMs
toward desired textual outcomes. This approach requires precision in articulating
creative intentions, an ability that differs from but complements traditional writing
skills.
Creative writing pedagogy may increasingly emphasise meta-literary awareness:
understanding genre conventions, narrative structures, and stylistic patterns at a
systematic level. While traditional writing instruction often teaches these elements
implicitly through practice, distant writing requires explicit knowledge of these patterns
to design prompts effectively. Students would need to analyse and articulate the
structural and stylistic features they wish to reproduce or transform in their work. In
short, distant writing requires close writing and reading.
The workshop model that has dominated creative writing education since the
mid-20th century may evolve toward what we might call a “design studio” approach.
Rather than focusing primarily on the critique of finished texts, workshops might
evaluate prompt strategies, compare multiple LLM-generated variations of the same
prompt, and collaboratively refine prompting techniques. This approach would
emphasise process over product, teaching students how to navigate the space of
narrative possibilities effectively.
Creative writing programs may need to incorporate elements from other
design-focused disciplines, such as architecture, game design, or user experience
design. Our age is the age of design, and these fields have developed sophisticated
methodologies for working with constraints, balancing functional and aesthetic
considerations, and designing experiences rather than objects, all skills relevant to the
distant writing paradigm.
The evaluation of student work may shift from assessing textual quality to
assessing design quality. Instructors would consider not just the final text but also the
effectiveness of the prompting strategy, the exploration of alternatives, and the
curation decisions made in selecting and refining LLM outputs. This represents a

(^17) I owe to Jessica Morley this remark.

significant departure from traditional assessment methods focused primarily on the
text itself.
Beyond these specific pedagogical shifts, distant writing raises broader
questions about the purpose of creative writing education: if LLMs can be used
successfully to produce good texts, what unique value remains for human-centred
creative writing education? Rather than competing with LLMs capabilities, creative
writing programs might focus on developing uniquely human strengths—conceptual
originality, design skills, cultural and ethical awareness, and critical judgment—that
complement rather than duplicate LLMs functions.
The democratising potential of distant writing also challenges the traditional
gatekeeping role of creative writing programs. WrAIting might enable more people
and empower those already enabled to care more about the novelty and the final result,
and less about the execution. If technical proficiency becomes less of a barrier to entry,
creative writing education might shift emphasis from teaching basic craft to cultivating
more advanced conceptual and critical skills. Of course, this democratisation is always
only partial, and its limits raise equity concerns in terms of exacerbating the digital
divide. Access to advanced LLM writing tools and the technical literacy required to use
them effectively are unequally distributed. Creative writing programs may need to
address these disparities to ensure that distant writing technologies expand rather than
limit access to literary creation.
Finally, creative writing education must grapple with the ethical dimensions of
distant writing. Students should understand the provenance of LLMs capabilities—
including the human-authored texts that form the training data—and develop
frameworks for using these tools responsibly. This might include discussions of
attribution, originality, cultural appropriation, and the changing value of creative
labour in an LLM-mediated environment.
Technological change often reveals implicit assumptions within educational
practices (Warwick 2018). The emergence of distant writing helps clarify assumptions
about authorship, creativity, and literary value embedded in traditional creative writing
pedagogy, providing an opportunity to reimagine these practices for a new
technological context.

11. Transforming Literary Criticism: New Paradigms for LLM-Generated Texts
The emergence of distant writing necessitates new critical approaches that adequately
address LLM-generated literature. Traditional literary criticism has developed around
assumptions of human authorship, intentionality, and cultural context that may not
apply straightforwardly to texts produced through human-AI collaboration.
To begin with, distant writing challenges the notion that a text’s meaning is
determined and controlled entirely and only by its author’s intentions. When
authorship is distributed between human designers and artificial executors (LLMs), the
locus of intention becomes ambiguous. Critics may adopt what could be termed design-
intention criticism , explicitly examining the link between human prompts and LLM-
generated texts rather than assuming traditional authorial intent.
Second, distant writing complicates biographical criticism. The biographical
details of human authors have traditionally informed interpretations of their work, but
how might we apply similar approaches to texts whose direct producer is an LLM?
Perhaps critics will develop methods for analysing the “biography” of LLMs—their
training data, architectural evolution, and deployment contexts—as relevant
interpretive frameworks. A related challenge will concern versioning, the availability of
drafts, annotations, and earlier, modified copies of a text. On the one hand, more could
be known if authors were to keep and share their prompts and a complete record of
the wrAIting process. On the other hand, the digital is incomparably more fragile and
transient than the analogue, so much could easily be lost.
Third, LLM-generated literature makes intertextuality more explicit and
pervasive, as LLMs are trained on vast corpora, and we have seen that they inevitably
echo patterns from their training data. Critics may need to develop new approaches to
identifying and interpreting these computational echoes, distinguishing between
patterns and features created by the designer, emerging from the wrAIting yet
endorsed by the designer, and mere artefacts of the training process.
Fourth, distant writing invites the reconsideration and expansion of reader
reception theory (Iser 1978, Jauss 1982). LLM-generated texts, shaped largely by
computational prompts, foreground the reader’s active role in meaning-making,
prompting renewed attention to reader agency, interpretative plurality, and
participatory forms of criticism. Reception theory can thus elucidate how readers

engage differently with LLM-created literature compared to traditional texts,
potentially shifting critical attention from authorial intention toward the diverse ways
readers interpret and interact with computationally generated narratives.
Finally, distant writing invites a meta-criticism that examines the critical
reception of LLM-generated literature itself. How do readers’ and critics’ conceptions
about LLMs authorship influence their evaluations? Do they apply different standards
to human-written versus LLM-generated texts? The reception of distant writing may
reveal as much about human attitudes toward creativity and technology as the texts
themselves. As (Hayles 2012) has argued about electronic literature, new textual forms
demand new critical approaches. Distant writing requires critics to develop
frameworks that can address the unique characteristics of LLM-generated texts while
maintaining the historical insights of traditional literary criticism.

12. Ethical Dimensions of Distant Writing: Questions of Attribution,
Originality, and Value
I have briefly touched upon the new ethical responsibilities of the author as a designer
and the ethical challenges that distant writing may generate, in terms of a new,
exacerbated digital divide. Other ethical implications of distant writing extend beyond
questions of attribution to encompass broader concerns about originality, intellectual
property, cultural production, and creative labour, raising profound philosophical
questions about technological mediation and responsibility. Here, the issues are well-
known and not unique to distant writing but to the whole world of LLM-generated
content, so I will only sketch them as a reminder.
By distributing creative agency among human designers, LLMs, and the
authors of the original training texts, distant writing complicates attribution and
accountability in literary production. This disruption has practical implications for
intellectual property regimes. Copyright law has traditionally protected the expression
of ideas rather than ideas themselves, but distant writing blurs this distinction by
making the human contribution primarily conceptual rather than expressive. (Boden

notes that our legal frameworks for creative work are still rooted in Romantic
conceptions of individual genius that may be inadequate for collaborative human-AI
creativity.
Beyond legal considerations, distant writing raises ethical questions about the
nature of creative work and its value. When writing becomes primarily a matter of
design rather than execution, how should we value different forms of creative labour?
Does the meta-author who designs prompts deserve the same recognition as the
traditional author who produces text directly? These questions echo debates in
conceptual art about whether the conception or execution of an artwork should be
privileged (Goldie and Schellekens 2007).
Training data complicates ethics further: since LLMs may use extensive
copyrighted texts, distant writing raises critical issues about proper compensation and
acknowledgment of original creators. The issue goes beyond the scope of this article,
but it is crucial. And it leads to a further concern: cultural appropriation. If prompt
designers generate texts in the style of authors from historically marginalised groups,
for example, without lived experience of the cultural contexts those authors represent,
this may constitute a form of digital colonialism, or at least cultural disrespect. Cultural
appropriation involves power imbalances that must be critically examined (Coombe
1998), and distant writing may exacerbate these imbalances by making cultural styles
more easily replicable.
Distant writing also raises sustainability concerns due to its reliance on
computationally intensive technologies. This is not an issue specific to distant writing,
but the significant energy consumption required by training LLMs in general poses
environmental challenges, prompting ethical considerations around responsible and
sustainable use of resources. Future discourse on distant writing should address these
impacts explicitly, balancing creative innovation with ecological responsibility.
Lastly, distant writing also raises existential questions about the value we assign
to human creativity itself. If LLMs can produce compelling literature based on human
prompts, does this diminish the special status we have traditionally assigned to human
creative expression? Or does it simply redirect human creativity toward design rather
than execution? These questions echo philosophical debates about authenticity and
meaning in an increasingly technological world (Taylor 1991).
All these questions will become increasingly pressing as distant writing
becomes an ordinary experience. The sooner we deal with them, the better.

13. Publishing in the Age of LLMs: Industry Transformations and New
Business Models
The publishing industry stands at the threshold of significant transformation as distant
writing becomes more prevalent, continuing historical patterns of technological
disruption and adaptation within publishing practices (Striphas 2009, Bhaskar 2013).
Traditional publishing models have been built around identifying, developing, and
marketing the work of human authors, individually or collectively, but distant writing
introduces new possibilities and challenges that may fundamentally reshape the
industry.
It is true that distant writing potentially democratises (in the computer science
not the political sense of the word) literary production by lowering barriers to entry.
Yet, this democratisation could lead to an explosion of new content, challenging
publishers’ traditional gatekeeping, selection processes, quality checks, and archival
roles. Publishers may need to develop new criteria for evaluating the quality and
marketability of texts produced through distant writing. This may lead to a significant
shift in the economics of publishing. Traditional royalty models assume one or more
authors, who receive compensation based on sales. Distant writing complicates this
model by introducing multiple stakeholders: prompt designers, LLM developers,
potentially the creators of works used in training data, and potentially readers who may
co-design the new texts (see below). Publishers may need to develop more complex
compensation models for these distributed contributions (Elkin-Koren 2017). Still part
of the economic transformation is that the text production timeline will accelerate
significantly. While traditional publishing involves lengthy processes of drafting,
editing, and revision, distant writing can potentially generate polished manuscripts
more quickly. This acceleration may pressure publishers to streamline their editorial
and production processes or risk being outpaced by more agile competitors leveraging
LLM capabilities.
Distant writing could foster customised publishing, even allowing novels to be
rapidly generated or revised based on reader preferences, similar to TV series
adaptations. In that case, publishers might offer personalised versions of texts tailored
to individual readers’ preferences or needs. This customisation could transform
reading from a mass consumption activity to a more individualised experience (Murray

2012). Agile and on-demand publications may become more common. The roles of
editors and copywriters may shift from direct textual revision towards designing and
optimising effective prompts, and carefully curating LLM-generated narratives.
Editors may assume responsibilities akin to film producers, managing and overseeing
the overall creative process rather than engaging primarily in manuscript correction.
This shift would require editors to develop new skills in prompt engineering and LLM
management. This is far from speculative. In 2025, Springer Nature begun offering
editorial roles to Subject Matter Experts (SME) to curate Human-Machine
Collaboration Books, see Table 2.

SMEs curate relevant research papers.
Our AI tool condenses these into Research Highlights.
SMEs organize them into a structure guided by you.
Your Contributions:

Define the book’s topic and table of contents.
Edit the TOC and finalize article selection.
Fact-check Research Highlights and add 2-page introduction.
Benefits:

Greater Reach: Broad institutional access.
Efficiency: Less time and effort for publication.
Speed: Faster production timeline.
Technology Exposure: Insights into cutting-edge AI and human creativity.
Table 2 Example of production of a Human-Machine Collaboration Book
Finally, distant writing may give rise to entirely new publishing business models.
Publishers might offer subscription services for access to LLM writing tools and
services with premium prompts, create marketplaces for effective prompts rather than
finished texts, or develop platforms that allow readers to generate their own literature
based on specific parameters. These new models would represent a significant
departure from traditional publishing approaches centred on the production and
distribution of fixed texts.
The publishing industry has historically adapted to technological changes, from
the printing press to e-books, by transforming its business models and value

proposition (Thompson 2021). Distant writing represents a major technological shift,
challenging publishers to reconsider how they create and deliver value in an era when
the production of written content is increasingly automated and distributed.

14. WrAIting and the Future of Literary Production
Looking forward, distant writing represents not just a novel approach to literary
creation but potentially a fundamental shift in how written communication is produced
more generally. One day, people may wonder how past generations could write all their
texts without LLMs doing most of the work. They will look at writing a bit like we
think about sewing our own clothes or growing our own vegetables: doable, but not
an ordinary practice.
Close writing might become an exceptional skill rather than the norm,
comparable to home-baking bread in a world of mass-produced goods. In the future,
most textual production may involve human-AI collaboration, with humans
establishing parameters and LLMs executing the actual writing. Such a shift would not
necessarily diminish human creativity but might redirect it toward design rather than
execution, with significant consequences which one can only begin to imagine. We live
in a culture based on close reading and close writing. If writing becomes wrAIting,
hence primarily a matter of design rather than execution, what will happen to the
intimate relationship between language and thought that has characterised human
expression? If everybody is an architect and nobody a mason, how will this affect our
relationship with the processes linked to reflection, understanding, or creative
thinking? For millennia, we have become accustomed to thinking through writing.
What will happen when we think through wrAIting? These questions highlight the
need for critical engagement with distant writing not merely as a technological
innovation but as a cultural and philosophical transformation with far-reaching
implications for understanding human expression. Let me close this section by quoting
Flusser again:
The new computer codes have made us all illiterate again. A new literate cast
has arisen. For most of us, the new writing (computer programmes) is suffused
with that kind of mystery that surrounded alphabetic writing before the

invention of print. [...] We have to try to use a typographic way of thinking to
get to grips with post-typographic “writing”. (Flusser 2011), pp. 55- 56.
15. The Fourth Revolution: Decentralising Human Information Production
Distant writing can be positioned within a broader historical context, what I have
described in the past as a fourth revolution (Floridi 2014), following the Copernican,
Darwinian, and Freudian revolutions that displaced humanity from the centre of
physical, biological, and mental spaces. This fourth revolution, which I attribute to
Turing, displaces us from the centre of the information space (infosphere). This
framing locates distant writing within a larger philosophical shift away from
anthropocentric conceptions of information production. As previous revolutions
challenged human centrality in various domains, distant writing challenges the
assumption that meaningful textual creation must be exclusively human. Drawing on
the philosophy of information (Floridi 2010), we can understand distant writing as part
of a broader reconfiguration of human-information relationships. In this
reconfiguration, humans shift from being the sole producers of meaningful
information to being mostly designers, curators, and interpreters within an increasingly
complex informational ecosystem that includes artificial agents. This perspective
suggests that distant writing is not just a technological development but part of a
fundamental philosophical reorientation that requires one to reconsider traditional
assumptions about human exceptionalism in creative production. It invites us to
explore “distributed morality” (Floridi 2013, 2016) and “distributed creativity” (Literat
and Glaveanu 2018, Bruno and Canina 2020, Mejia, D'Ippolito, and Kajikawa 2021),
where meaningful creation emerges from networks of human and non-human agents
rather than from individual human minds alone.
16. Conclusion: The Bounds and Horizons of Distant Writing
Distant writing represents a methodological innovation in literary production and a
conceptual challenge to traditional understandings of authorship, creativity, and
narrative. Positioning the human as a designer rather than a direct producer of text
reconfigures the creative process along lines previously more familiar in fields like
architecture or fashion design than in literature. This reconfiguration reveals several

valuable new insights about narrative and literary creation. It shows the boundless but
constrained nature of narrative possibilities, the isotropy of narrative space, and the
connected network structure of fiction. It challenges us to think differently about
narratives, stylistic signatures, and the future of literary production. Distant writing
should not be perceived as replacing traditional writing but rather as expanding creative
possibilities. It provides novel tools for navigating narrative spaces, linking diverse
literary universes, and democratising literary production. As with all significant
innovations, it inevitably raises complex questions about authorship, creative agency,
ownership, and the evolving human-technological relationship. However, these
challenges, when they do not translate into illegal practices, should be welcomed as
opportunities to understand human nature and predicament more deeply and refine
and expand human ambitions and creativity.
As we continue exploring the potential of distant writing, we should approach
it not with uncritical enthusiasm or reflexive scepticism, but with thoughtful
engagement that recognises its possibilities and limitations. The future of literature
may involve increasing collaboration between human designers and LLMs, but the
value of this collaboration will depend on our ability to develop frameworks that
preserve meaningful human creative agency while embracing the expanded
possibilities that distant writing offers. In the end, distant writing invites us to
reconsider not just how we produce literature but what literature is and might
become—a reconsideration that may prove as significant for our understanding of
human creativity as Moretti’s distant reading has been for our understanding of literary
analysis. Some good news is certain: there is plenty of exciting intellectual work to
make the most of and understand distant writing.

Acknowledgements
I am very grateful to Emmie Hine, Jessica Morley, and Claudio Novelli for our many
conversations on this topic and their insightful comments about previous versions of
this article. They keep me on my toes. Once I shared a draft of the article online
Hannes Bajohr kindly called my attention to an essay of his (now in the references)
that I had missed in the first version of this article, and Scott Newstok shared his
helpful work on close reading and several references to distant writing, which are also

now referred in this version (he also caught several typos). This is the beauty of the
web. Kia Nobre is the reason I could write it. All remaining mistakes are mine.