Split Fiction: Gaming, Authorship, and Corporate Extraction in the Age of AI

Abstract

This article examines Split Fiction, a cooperative video game that engages with themes of authorship, creativity, and artificial intelligence in the digital age. The game presents aspiring authors whose creative ideas are extracted by a corporate machine—a metaphor for contemporary generative AI systems. Through its mandatory two-player cooperative mechanics and genre-shifting gameplay, Split Fiction explores tensions between human creativity and automated generation, individual authorship and corporate extraction, and procedural rhetoric versus narrative meaning. We analyze how the game’s mechanical variety, intertextual references, and meta-narrative structure comment on the current landscape of AI in creative industries, particularly as director Josef Fares’s ambivalent statements about AI complicate straightforward readings of the work as purely anti-AI critique. The game ultimately offers a nuanced exploration of creative labor futures in an age where the boundaries between human and machine authorship grow increasingly uncertain.

1. Introduction

Split Fiction (Hazelight Studios 2025) opens with a simple scene: aspiring authors enter a giant room with a huge whirring machine after agreeing to become published. They all suit up and are encased in glowing, embryonic bubbles as part of a strange corporate technology experiment. The exact ramifications of this experiment are not revealed until much later in the sequence. We also meet two young writers who are excited about becoming published and see this trial as an opportunity to be published. However, the game’s engagement with authorship is mostly not focused on the literary. Instead, the work begins by shifting between two genres—fantasy and science fiction—with changing game mechanics that rely on both players communicating and understanding each other’s role in solving any given puzzle. The gameplay itself references and uses mechanics from previous game franchises and genres. Its experimental approach of cooperative mechanics and its conceit of a meta-story highlight its features as a work of interactive narrative, where computational processes are necessary to its meaning. This layering of meaning of gameplay and narrative puts the technologies of play at the center, as Split Fiction ultimately offers games within the frame story game and suggests that within the narrative, those games are not designed but instead generated. This is visible through the premise of immersive story worlds drawn from the authors’ minds, which readily evokes generative AI: the worlds themselves are pulled from their minds and fully realized (rendered, even, mostly in 3D) in a way that their imaginations might not be capable of. Their “ideas” have been given graphical and procedural form that an emerging narrative designer might envy: while the writers are centered, the playable outputs are clearly not their work. This wish fulfillment foretells a future of generative AI that sees creativity being unlocked by AI. And yet the game itself also works as a critique of how the AI technology today has an immediate and significant impact on ownership and control of creative works. AI presents an existential challenge to creative industries, promising both to replace and to devalue entire careers and the outputs of creative professionals. The game’s plot seems to reflect on the current usage of generative AI and the resulting unstable relationship to human creative output and ownership. This relationship is highlighted through the game’s tension between homage to previous works and reuse of previous genre conventions and tropes. Director Josef Fares’s nuanced and sympathetic position on AI complicates surface readings of the work’s message: it is not simply a story about corporate AI profiting from the stolen ideas of creators, but one that explores the tensions around this technology experienced by those of us working broadly in the space of computational narratives.

2. Materials and Methods

This article employs close play methodology (Chang and Welsh 2025) to analyze Split Fiction (Hazelight Studios 2025), examining both its ludic mechanics and narrative structures through sustained engagement with the game’s cooperative gameplay. Our analysis draws upon the complete playthrough of Split Fiction alongside a curated corpus of critical reception materials, including professional reviews, developer interviews, and community discourse published within six months of the game’s March 2025 release. Close play, as a methodological approach, extends close reading practices to interactive media by attending to the relationship between procedural systems, player experience, and textual meaning. Following the model of autoethnographic game studies exemplified by Wardrip-Fruin’s sustained engagement with Animal Crossing (Wardrip-Fruin 2025), we acknowledge that our close play was shaped by the social dynamics of our collaboration. The two authors played through Split Fiction together as the mandatory co-op partners, bringing different relationships to the genres central to the game: one more versed in science fiction games and narratives, the other in fantasy. This dynamic mirrored, perhaps too neatly, the relationship between Mio and Zoe themselves. We played Split Fiction as much as possible without consuming any external reference media that would influence our play or perception of the game, choosing to avoid streamer content that would bring in the additional dynamics of another pair’s experience of the game. Our negotiations over gameplay priorities when mechanics demanded coordination inevitably informed our reading of the game’s themes around collaborative authorship. We examine how the game’s mandatory two-player cooperative mechanics, constantly shifting genre conventions, and narrative framing work together to articulate themes of authorship and creative extraction. The selected interviews and reviews provide crucial context for understanding both the game’s reception within broader debates about generative AI and the tensions between authorial intent and textual interpretation, particularly regarding director Josef Fares’ public statements about AI technology in game development.

3. Contextualizing Split Fiction

Split Fiction was released in March 2025 as the third game from Hazelight studios, a Swedish game development company founded by film director Josef Fares and known for cooperative games. The studio’s 2021 title, It Takes Two, won Game of the Year from The Game Award and was nominated for a number of other awards, particularly for multiplayer, which has become the studio’s distinguishing emphasis: while cooperative play is present in a number of titles (particularly in the console world, which often includes couch co-op play as part of the appeal for families), it is relatively rare for narrative games, and even more unusual for a game to require two player co-op for any progression. This trend in video games is reminiscent of a mode of cooperative play that has emerged in board games and particularly been propelled forward by the pandemic: the two-player cooperative board game, marketed initially by Czech Games Edition with the release of Codenames: Duet (Chvátil and Eaton 2017) as “duet” play and later adopted by other game makers, is targeted at couples looking for play that is balanced well for two—many larger multiplayer titles are by contrast not neatly playable by two, and while the “duel” format is also popular, we at least can attest that there is value in a non-combative game night. In It Takes Two (Hazelight Studios 2021), the relationship therapy metaphor is made literal: the game has rightly attracted criticism for its narrative emphasis on a couple undergoing divorce, and the suggestion that the rift between the two members of the couple can be resolved through cooperation and a literal overcoming of obstacles. As one reviewer noted, this dissonance detracts from the play: “In theory, you’re supposed to be following May and Cody on a sweet, endearing journey as they relearn working together as a team and (spoiler) eventually reconcile. In practice, the game presents the couple as incredibly selfish people who absolutely should get a divorce, and who should, at the very least, seriously rethink their approach to parenting” (Thurm 2022). Josef Fares has spoken to this aspect of the studio’s co-op games as presenting a relationship test, in and of itself, noting in an interview:

We realized quite early, when we did some early testing, that some couples who came to test the game, we were like, “Oh shit, they need to work on their collaboration.” And I still say today that if you want to see where your relationship is, you should play our games. Because it’s a good way to see if your communication is actually good. Because when we design the gameplay and the characters, it’s key that you can communicate.

(Franklin II 2025)

Split Fiction continues the studio’s tradition of intentionally designed co-op between two characters with asymmetrical abilities that complement one another for the solving of a wide range of puzzles: those abilities, and the metaphors driving them, change as the game progresses. However, it presents two characters with a less dramatic conflict, and in doing so places much less emphasis on resolving a relationship tension and more on the external threats posed by both the characters’ own stories and the company trying to possess them.

The game’s two protagonists embody this tension between aspiration and exploitation. Zoe, a young woman grieving the loss of her sister, writes primarily in fantasy—her stories populated by knights, dragons, and quests that capture the innocent childhood she shared with her sister before the tragic accident. Mio, more reticent and guarded, writes science fiction that imagines futures dominated by technology and corporate power. Their initial antagonism—each dismissive of the other’s genre—gives way to reluctant collaboration as they realize the machine has trapped them together. Both are presented as writers who have struggled to complete or publish their work, making them vulnerable to the games’ villain, J.D. Rader, who is founder and CEO of the tech megacorp, Rader Publishing. They promise to take promising writer’s ideas and “publish” them.

Mio and Zoe are both deeply invested in genre tropes, and their dislike of one another’s storytelling styles is presented as a dynamic of conflict to fill the gap in a way that feels particularly manufactured: as one reviewer critiques, “The central conflict between the two playable protagonists revolves around a bizarre “jocks vs. nerds”-style dynamic between sci-fi and fantasy writers” (Volk 2025). This also gets to the heart of one of the most disconcerting choices in the work, which is to suggest that the characters are unpublished writers, but present their works entirely through play:

Both Mio and Zoe are supposed to write books, but it feels like no one involved is aware of what a book is. Most of their “ideas” we play through in Split Fiction are simply a series of settings or minigames, or surface-level pastiches of existing popular properties … these worlds lack any of the details real-life writers would likely bring to them—for example, as far as I can tell, no character is named in any of these fictional worlds. No other people seem to exist within the stories. The places are not named.

(Volk 2025)

Yet the framing of authorship, thinly “written” though these worlds might be, is at the heart of Split Fiction: however, the authorship in question is not that of text, but that of the games and interactive narratives broadly.

Split Fiction’s mandatory co-op gameplay presents an interesting take on the current generative AI prompt-based workflow. Not even supporting an AI player argues strongly that human interaction is critical for the game’s meaning and success. The world that was created required not just the ideas of the participants in the study, but their active engagement. The game takes this even further: games are only existent when played by players, and in this required co-op genre, they are further only possible when three human roles communicate and collaborate, with two (the players) at the same time. This suggests that generative AI, as powerful and as impressive as it may be, is ultimately only as valuable as the human experiences it provokes, and that the massive investment in data centers, technologies and companies currently dominating the industry of generative AI will ultimately require justification and value that can only be proven through human interactions. While current collaboration with generative AI focuses purely on a traditional interactivity feedback loop described by Chris Crawford’s “Speak,” “think,” and “act” stages (Crawford 1982), it is clear that the dynamic interplay of multiple humans and (multiple) AI processes will be far more interesting that today’s prompt-based interactions.

These dynamics of human-machine play connect to broader questions in co-creativity research about how creative responsibility is distributed in computer-aided processes. Mollick’s recent work on “co-intelligence” frames the emerging relationship between humans and AI as one requiring new modes of collaboration rather than simple tool use (Mollick 2024). Haase and Pokutta’s framework for human-AI co-creativity identifies multiple levels of interaction, from AI as “digital pen” to AI as genuine “co-creator,” noting that generative AI systems now “contribute actively, demonstrating autonomous creativity in producing novel and valuable outcomes” (Haase and Pokutta 2024). Under this framework, Split Fiction’s pods literalize what they call “task-divided” collaboration—humans provide ideas, machines provide execution—while the narrative frames this division as exploitation rather than partnership. This tension between co-creative potential and extractive reality sits at the heart of current debates about generative AI in creative industries.

These questions of whose vision a “world” represents are particularly well-represented in technology today by the efforts of the Google DeepMind Genie project. Genie is marketed as an AI “world” model intended to work not unlike the pods of Split Fiction: “with world models, you give them a prompt and they generate a space that you can move around in like you would in a video game, but instead of the world being handcrafted with 3D assets, it’s all being generated with AI” (Peters 2025). Two versions have been demonstrated to the public in 2025: one reporter covering the most recent iteration noted that this suggests “AI-generated video games might become a reality faster than previously thought” (Kan 2025). For Mio and Zoe, the pods are the equivalent of a future version of Genie, an unexplained, apparently inhuman technology that takes their ideas and turns them into games—but whose games are these, really? Are we in Mio and Zoe’s stories as we play, or are we truly in their imaginations as processed by an imagined iteration of generative and agentic AI—one that gets to the heart of what multimodal authorship might become?

4. The Game as Genre Survey

Aspirational authorship frames Mio and Zoe’s quest, and with apologies for the spoilers, the game ends with the two authors collaboratively writing a book about their experience after overcoming the evils of the corporate fiction empire at the center of the narrative. However, this moment of bookishness is a rare one within the game’s own conceits, as Split Fiction is essentially a survey of the genres of play that have developed over the last several decades. While genres are themselves a contested and potentially inaccurate method of interrogating games (Apperley 2006), they are used by game critics and publishers to help set expectations and to otherwise “classify” games for the general public. Jessica Cogswell of GameSpot observes that “several tools and powers were so thoughtfully designed they could have easily been the crux of an entire game” (Cogswell 2025). The sheer mechanical variety becomes overwhelming; as one Inverse reviewer notes, the game features “a near constant stream of innovation and genius” that leaves players questioning “how Hazelight, let alone any developer making games today, will ever top” this level of invention (Dowd 2025). Through its mashup of different genre mechanics, the game invites players to cooperate and engage in co-creation while having no control over the narrative whatsoever. The tension between narrative railroading and mechanical variety exemplifies what Sicart (Sicart 2011) identified as a fundamental limitation of procedural rhetoric: rules as the container of meaning rather than playing as an act of meaning-making. The very shifting of the mechanics undermines a procedural reading of the piece, where the mechanic switching may even undermine the proceduralist reading. At the mechanical level, Split Fiction offers several assistive settings—more resilience in boss fights, easier targeting, and other features for players more interested in following the larger narrative through. The variation of the difficulty settings may not be worthy of procedural rhetoric analysis, but speaks to the pragmatic goals of requiring two players of potentially different interests and skills to collaborate in order to progress.

5. Anxiety over Authorship

At the heart of Split Fiction is an anxiety over authorship, particularly striking given that few of the game’s singular narratives or moments feel particularly original. Instead, it traffics in familiar tropes. This derivative quality reflects what Consalvo describes as the intertextual nature of games, where meaning emerges through references to other texts and shared cultural knowledge (Consalvo 2003). The two lead characters represent genre fiction authors in the most easily dismissed way. As Tessa Kaur argues in TheGamer, “For a game that’s supposedly about writers fighting to hold on to their original ideas, very little of Split Fiction’s ideas, at least in terms of worldbuilding and plot, are interesting or original. In fact, the game seems to reference other video games far more than it creates original ideas or even reference literary work” (Kaur 2025). This critique cuts to the core contradiction: “Maybe these writers aren’t getting published because all their work rips off other existing franchises with barely any changes” (Kaur 2025). This disconnect between narrative ambition and ludic execution reflects what Klevjer identifies as the fundamental tension in how games communicate—through the “ergodic” work of gameplay versus the “interpretive” work of cutscenes and narrative framing (Klevjer 2003). These plots serve as a narrative layer on top of intense, combat-driven or twitch-based mechanics of play. The frame story is more about the rationale behind the setting and boss than any true narrative actions. The stories from each author have things in common with fan fiction, featuring simple heroic quest plots and metaphors from their own lives exaggerated to create emotional meaning.

The derivative quality of Mio and Zoe’s narrative fragments—populated by familiar tropes, archetypal characters, and genre conventions—resonates with the creative practices of fan fiction communities. Jenkins’s foundational work on “textual poaching” describes how fans appropriate and transform existing narratives, creating new works that exist in productive tension with their sources (Jenkins 2013). Split Fiction’s protagonists operate in a similar mode: their “ideas” are less original creations than recombinations of beloved genre elements. Yet where fan fiction communities have historically operated outside commercial structures—or in uneasy negotiation with them—Rader Publishing’s pods represent the full commercialization of this creative mode. The parallel to contemporary platforms is striking: just as Wattpad and Kindle Worlds promised aspiring writers visibility while harvesting their labor, the game’s fiction-extraction machine offers publication in exchange for creative autonomy.

In their study of the relationship between fans and media industry companies, Stanfill (2019) argues that media industries have sought the “domestication of fandom.” This has historically played out through efforts to invite fans to consume and even participate in media franchises through ways that are useful to the industry in turn—discouraging transformative practices such as fan fiction circulated on non-industry platforms like Archive of Our Own, while encouraging participation in sponsored fan art contests and other spaces where fan labor serves the promotion and amplification of the original property. Split Fiction’s imaginary platform similarly creates a space in which the labor of the aspiring author—and both Mio and Zoe can be read as fan authors—generates content from which the media industry profits. This dynamic echoes how game industries specifically have benefited from substantial fan labor, with one of the most notorious examples being the Roblox platform, which has faced criticism for its exploitation of players as a sandbox game that relies upon user-generated content for its aesthetics and appeal (Nelson 2023). The derision fan created-material, including fan fiction, commonly faces has played a role in enabling this exploitation.

Yet critics remain divided on whether this derivative quality undermines or reinforces the game’s themes. While some see the references as organic to the premise, others find them hypocritical. Giovanni Colantonio of Digital Trends delivers a meta-critical observation: “This is a game about books that could not be less interested in actual literature. We’re supposed to be playing through visualizations of Zoe and Mio’s novels, but those levels only ever speak in the language of movies and video games” (Colantonio 2025). He poses a pointed question: “How many books have you read where the heroes must dance with a monkey king by doing a Simon Says minigame?” (Colantonio 2025). The game’s performative creativity—its own relentless mechanical invention and polish—ultimately undermines its shallow engagement with writing craft. The fragments within the game demonstrate just how superficial the narrative overtones in many game genres remain.

Split Fiction is scheduled to be adapted into a film—a strange proposition given that it already presents as a story about game structures. As reported by Digital Trends, “Hazelight is reteaming with Story Kitchen—the production studio that helped option It Takes Two as a movie for Amazon MGM Studios—to adapt Split Fiction as a film” with the companies “reportedly already bringing together a director, screenwriters, and cast members as a package deal” (Marnell 2025). The challenge of this adaptation speaks to what Egliston identifies as the fundamental incompatibility between ludic and cinematic forms—games create meaning through “transtextual” mechanics that cannot be directly translated to linear media (Egliston 2019). The experience is fundamentally about game genres and how we experience them. In adaptation, the narrative would presumably still center on two writers aspiring to write fiction, but in the hands of a thoughtful director, the work would likely move from the tropes of video game mechanics to the tropes of film, making the final results potentially unrecognizable. This “unadaptability” is part of what makes Split Fiction interesting as a game narrative, and also challenging to read simply as a narrative about authorship.

The most substantive engagement with authorship appears in a Notebook side story, where the game briefly explores the creative process itself. As TheGamer notes, this sequence “plays out like a draft of a story” and “feels like a discarded idea lodged inside a writer’s brain, prodded at occasionally like a sore tooth” (TheGamer Staff 2025b). Critics consistently praise this moment as the singular instance where the game authentically engages with what it means to write, as opposed to the broader discourse of the game which conflates authorship with world-building.

6. The Corporate Machine and Generative AI

In examining Split Fiction reviews and discourse, we argue that the game’s narrative structures and mechanics enable it to speak particularly strongly to resistance against corporate machines that extract individual authorship. It is difficult not to read the construct at the center of Split Fiction as a metaphor for generative AI—and the metaphor is not subtle. Director Josef Fares himself confirmed to Inverse that “There’s definitely a connection to that [generative AI]” and described the pods as “symbolism for generative AI” (Liao 2025).

The opening gameplay sequence shows all the writers hooked into a large centralized machine, suspended in bubbles where they play through or live out their constructed narratives. Our two leads—one representing science fiction, one representing fantasy—end up in the same bubble. Their narrative constructions collide, creating interesting tensions from emergent contradictions and possibilities, even as their narratives remain trope-filled, cliche-laden, and intertextual with homages to both games and popular fiction series. Many critics read this as urgent cultural commentary. As one AIPT review argues, Split Fiction presents a “timely story of humanity vs. technology; creativity vs. uninspired slop” (Manchester 2025). IGN frames it even more forcefully as ‘two proud middle fingers to the conga line of entitled tech bros who’ve decided art is a commodity that can be scraped, stolen, and repackaged by their plagiarism engines” (Reilly 2025). This framing holds echoes of the designer-forward, anti-AI, stance that is becoming common in the games industry, particularly in the wake of increased layoffs and rising concerns about the future of creative labor.

However, this straightforward anti-AI reading becomes complicated by Fares’ own statements about artificial intelligence in development. In interviews published before the game’s release, Fares told VGC: ‘If it’s part of the industry we should see how to implement it to see how we get better games. I can understand the fact that some people could lose their jobs but that goes for every new technology” (Middler 2025). He elaborated to Inverse: ‘At the end of the day, it all goes down to the vision and then the passion, and if AI becomes a great tool or not, it’s only going to hopefully help us make even better games” (Liao 2025). These pro-adaptation stances generated significant negative community reaction, with Kotaku noting, ‘I would have preferred a more robust rejection of AI from Fares” given the game’s themes (Shepard 2025). While the technology corporation is clearly the villain of the game, this framing makes it harder to easily read the game as a critique—particularly given the elevation of an AI-esque generated play at its core.

Fares’s ambivalent position reflects broader divisions within the games industry over AI adoption. The 2024–2025 SAG-AFTRA video game strike, which lasted nearly eleven months, placed AI protections at the center of labor disputes, with voice actors and motion capture performers ultimately securing provisions requiring consent and fair compensation for digital replicas of their performances (SAG-AFTRA 2025). Meanwhile, indie developers have increasingly positioned themselves against generative AI: studios like Strange Scaffold and Necrosoft Games have publicly rejected AI tools, with Strange Scaffold’s creative director Xalavier Nelson Jr. dismissing industry claims that “every game company is now using AI” as “normalization bulls**t” (Woods 2025). The Game Developers Conference 2025 State of the Industry report found that 30% of developers now believe AI is actively harming the industry—a twelve-point increase from the previous year (Game Developers Conference 2025). Fares’s statement that job losses are inevitable with “every new technology” thus positions him closer to management rhetoric than to the growing labor movement within games, creating tension between Split Fiction’s apparent critique of corporate extraction and its creator’s expressed pragmatism about the tools enabling that extraction.

TheGamer’s post-launch critical analysis questions whether the AI metaphor succeeds at all. The article argues that “The machine never steals work—these women are writers, yet it’s not their prose that’s being put into a Large Language Model, but their ideas” and asks, “What does this have to do with the craft of writing?” The critique concludes: “Perhaps that’s not the case at all, but the alternative is that the writing is as confused as Fares’ opinions. And, to be fair, the writing’s pretty bad overall, so that’s entirely possible” (TheGamer Staff 2025a). As they navigate this juxtaposition, they become aware of the larger corporate overlord who represents the machine, extracting the creative energy of these individual writers to feed his system. It is difficult not to see echoes of recent AI lawsuits and the suspicions of digitization that have haunted the craft of fiction, especially in this current moment when questions of narrative ownership and control grow murkier with the pressing realities of model training on both corporate narratives and the work of unpublished writers who have provided their words for free to fuel the larger corporate machine.

Another reviewer, Trone Dowd, praised the game while noting the way the discussion of ideas hits particularly hard:

There’s a moment where a company investor shares concerns that the publisher may sacrifice the lives of its writers to this experimental tech, to which Rader flippantly tells an investor, “ideas are easy.” It’s a particularly memorable line that hits like a gut punch when delivered, especially when you know most of those in charge of funding and delivering our favorite games, movies, music, and TV to the masses share that exact mindset. Split Fiction directly addresses the single greatest threat facing creatives today, and it’s a wonderfully relevant game because of it.

(Dowd 2025)

However, that moment might instead be read as a valid critique of the two protagonists, both presented as emerging writers full of ideas with lots of unfinished stories. Their pursuit of the pods is literally the pursuit of generative AI to fill in the gaps in their own imaginations—throughout, there is a continual suggestion that the authors don’t recognize their own worlds and stories. The experiences they have entered, that the player is exploring, represent an artistic endeavor that goes beyond the text.

7. Splitting Mechanics

Like films, games are inevitably self-referential. They not only build on previous games for their mechanics, but often require player familiarity with them in order to appreciate their variations and evolutions. The underlying ability for a computational work to convey meaning through not only its surface level representations but through its systems and processes is what Wardrip-Fruin noted in games and systems which have varying degrees of observability (Wardrip-Fruin 2009). While Split Fiction is not a game with complex emergent behavior, the interplay of two players and the mechanics designed to require them to work together is a good example of the underlying innovation within the work and something that moves the game beyond some of the mechanics that are recruited to achieve its effects. As Apperley argues, game genres are less stable taxonomies than ‘assemblages of conventions’ that developers and players recognize and reconfigure (Apperley 2006). To create a platformer is to build a game which assumes the players have some experience with previous games in the genre. The developers know the constraints, the opportunities within the mechanic. While the mechanics of platformers, fighting games and racing games may be familiar and well trodden, many of these mechanics require additional thinking in order to successfully achieve two player game play. There are few traditional console games that include, much less require, asymmetric coop, with most favoring a symmetric split screen approach (Mario Kart (Nintendo EAD 1992), Halo (Bungie 2001) series) and party games in the WarioWare (Nintendo R&D1) series offering multiplayer challenges that are cooperative and asymmetric in one game mode. This scarcity reflects what Rocha et al. identify as the design challenge of cooperative mechanics—balancing individual agency with shared goals (Rocha et al. 2008). Seif El-Nasr et al. further note that successful cooperative games must negotiate between ‘enforced’ and ‘encouraged’ cooperation, a tension Split Fiction exploits through its mandatory two-player structure (Seif El-Nasr et al. 2010).

In Split Fiction, the developer’s choice to require a second player makes the collaborative mechanics essential, just as in the previously discussed “duet” 2-player board games. The core platformer mechanics of dashing and jumping and navigating 3D worlds are only present in a subset of the levels: players must quickly learn and apply a series of new abilities based on each fictional world, and each character’s skills and role are distinct in each segment. This constant novelty distinguishes Split Fiction from traditional games that refine a single mechanic through environmental puzzle design; here, mechanical mastery is deliberately impossible, replaced by adaptability and communication between players. For Split Fiction, it does make sense when most of its mechanics are familiar from previous games to be able to avoid excessive tutorials. There are some genuinely innovative combinations of the mechanics and the two players having to work together even as many of the core game mechanics are as familiar as the fictional stories that they inhabit.

The meaning behind mechanics can be itself a complicated topic. Wardrip-Fruin discusses how the interpretation of the mechanics (what he calls “operational logics”) in games like Pac Man (Namco 1980) can be a lens into how games requires a different type of interpretation on the part of players (Wardrip-Fruin 2020). Yet as we mentioned previously, Split Fiction’s constant mechanical shifting could be itself an example of what Bogost calls ’procedural rhetoric’—the idea that games convey messages through their rule systems, and that the arguments made by the mechanisms of play can be as (or more) critical than the surface narrative and aesthetics (Bogost 2007). Instead of relying on meaning through the traditional sustained engagement with a coherent set of mechanics, Split Fiction relies on continual novelty to convey the variety of ideas and creativity of its primary protagonists, and expresses the contradictions and possibilities of their changing imaginary worlds through these shifts.

8. Notebook

There are a few sequences within that novel play that are innovative in both the stylistic departures from the main storyline chapters and in their willingness to play with expectations. But like other sections, the “Notebook” side story primarily depicts the act of creation and the act of creating choices in a way that evokes The Stanley Parable (Wreden and Pugh 2013) and Undertale (Fox 2015). It depicts an uncertain narrator–author exploring an idea for a story with erasures and changes to key elements such as the player’s weapons or the enemies. Ultimately, though, the story is more of a game than a narrative, with action (including enemies and hazards) dominating the gameplay over dialogue or character development. One section allows players to choose between options by “voting” for them, not allowing them to proceed until both players agree by standing next to an option with a bar filling up. For instance, in one scene, the text “But a ___ has captured the prince!” appears, with “DEMON,” “CRAB” and “DUCK” appearing below for players to vote on by moving their characters below the words. The game will not progress until both sides have been filled by the player standing there long enough, mirroring other gates in the game where both players must simultaneously activate a device or mechanic to proceed to the next segment. At the end of the sequence, the pencil rendering technique is kept while the split screen is restored along with a 3D world as the two characters bring the prince (who is now sitting on the back of one of the characters) to a sailing ship.

Anthropic has recently presented a new prototype of one vision for creating user interfaces that resembles the types of in-process creation and exploration depicted in this sequence, called “Imagine” (Anthropic 2025). Anthropic released it as a limited research preview, and users were able to prompt the creation of applications that were designed as user interface elements directly by a Large Language Model (LLM), which would change and update based on what the LLM believed would happen. This approach of just-in-time creation resembles the gameplay in the sequence and is a simpler and more near-term version of what Google Genie demonstrated (Kan 2025). While these may currently be the stuff of tech demos and venture capital fundraising, the labor futures they foretell promise the same erasure of designers, coders, and writers at the heart of Split Fiction.

9. The Battle Against Rader: God in the Machine

This metaphor becomes particularly revealing in the sequences that take place outside of the main narrative fragments—outside of the narrative fragments, I should say—that reveal how the system itself operates. As the machine that Mio and Zoe are plugged into starts to break down, the white man with brown hair and glasses standing in for every tech CEO ever declares “you are not destroying my machine” and jumps into it. Rader, in effect, joins Mio and Zoe in the bubble, in the Tron space, but brings to it this presumed mastery over the machine.

He emerges into the space not as a human wearing a Tron-esque suit of glowing lines like Mio and Zoe, but instead declares himself to be the one who’ll make the rules, proclaiming “I am God” as his glowing body pops into their space and starts to push them into other narrative fragments in which he can now mess with their very perception of reality. GameSpot notes that the game’s “finale saves its best gimmick for last” with a level that “makes you wonder what a whole game like this could have been like” (Cogswell 2025). This has consequences that play out in the game. Mio and Zoe perceive the same world but through the lens of fantasy and science fiction, each with different abilities to make use of and advance those juxtapositions. As Rader gets increasingly involved, we see his gold-outlined fiery visage towering over the landscapes through which they’re fighting. This gets at its most interesting and computational in the very Tron-esque, building blocks, Monument Valley-looking sequences in which they fight Rader directly. The Tron resonance extends beyond visual aesthetics to questions of power and creation. In Tron (1982) and its sequel, the digital world operates according to rules set by its creator-users, who appear as godlike figures to the programs inhabiting their systems. Flynn’s ability to manipulate the digital environment stems from his status as a “user”—a creator rather than a created being. Rader’s declaration that he is “God” thus invokes not merely megalomania but a specific genealogy of digital authorship: he claims the privileges of the creator-class, positioning Mio and Zoe as mere programs running within his system. That the players ultimately defeat Rader by exploiting glitches—system failures that escape his control—suggests that even totalizing creative authority has limits. Rader’s control over the system exemplifies what Fizek calls the “automated state of play”—when the game itself becomes an active agent that plays against or despite the human players (Fizek 2018). As this battle progresses, Rader demonstrates an increasing lack of vulnerability to their attacks but also control over the system, which starts to break down. In one memorable sequence, borders between different types of landscape move back and forth as the players try to navigate a safe world of lily pads versus giant terrain of hungry piranha versus Rader’s more technology-infused world.

Rader’s vision of himself as a video game boss is clearly informed by other media but also starts to get particularly meta near the end of the sequence as we become aware of the game as controlled. In different sequences, Mio is in the game literally on what appears at first to be a CRT screen, playing through a game in which Zoe can influence the game from the other side of the screen—a co-operative play mechanic that extends outside of that fictional reality. The next sequence moves to a handheld where we see on screen the physical hands of presumably Zoe controlling the screen and tilting it, using very familiar mechanics from iPad and tablet games that use rotation to govern the physics of objects on screen. This progression follows the trends of metanarratives that make us entirely aware of the game as an object—complicated, of course, by the fact that we’re already playing a game within the game as far as the narrative structures are concerned.

10. Conclusions

The player does defeat Rader when Mio and Zoe escape and break down the bubbles. The machine itself is destroyed, and all of the other authors who have been sucked into their computational representations of their narratives are freed from the machine as well. The final sequence shows Mio visiting her father’s grave with a copy of the co-authored book Split Fiction—presumably the written version of their experience destroying this particular company’s technology for good. Cogswell observes that Split Fiction “emphasizes that it takes humanity to create—that it takes the experiences that shape our lives and construct our subconscious to form stories” (Cogswell 2025). In doing so, they return from an experience of highly digital world-building and game tropes to the authorship of a traditional print book, which, of course, in our current landscape, becomes fuel for the generative AI machine.

Split Fiction offers an ideal narrative of authorship anxiety for our times, in which the struggles to regain control of one’s text—to literally be the owner of one’s narrative—play out through a series of game mechanics challenges requiring cooperation between two authors of divergent, dissonant genres. As Justin Clark of Slant Magazine captures the game’s contradictions: “Split Fiction is hokey, muddled, and needlessly self-defeating. It’s also lively, inventive, and so earnest that it’s hard to be mad at it for long. These aren’t opposing forces that tear Hazelight’s latest apart; the clumsiness is inseparable from the delight” (Clark 2025). The game’s cultural significance may lie precisely in these contradictions rather than its coherence—generating sustained critical debate about creative borrowing, authorship, and the nature of originality in interactive media at a moment when these questions have never been more urgent.

The game itself features a group of unpublished writers, all represented as people who’ve never managed to bring their work fully into the world, willing to literally sign away their autonomy in the hopes of having their unfinished ideas in “print.” They’re essentially feeding those ideas into a massive prompt machine that transforms fragments of imagination into playable experiences. In that sense, the machine can be read simultaneously as a metaphor for the thrill of exploring creative possibilities through AI and as an indictment of how those same mechanisms can consume and control artistic expression. This ties into broader concerns in the field over open source, authorship, and games, particularly as the encroaching capacities of AI technologies to replicate any property raise fundamental questions about the ownership of ideas and the capacity of even large companies to maintain agency over one’s creative work. In late 2025, that is being tested thanks to the release of OpenAI’s Sora 2, which initially launched with a controversial “opt-out” approach to copyright—assuming permission unless creators filed removal requests—before quickly pivoting to an “opt-in” model after significant backlash from Hollywood studios and creators (Roberts 2025). The initial launch was nonetheless flooded with videos of grilled Pikachus and similar encroachments on Nintendo’s intellectual property, demonstrating how quickly AI tools can be weaponized against brand identity.

So perhaps the real commentary in Split Fiction isn’t about generative AI as a technology, but about the encroaching realities of corporate AI in the games industry and narrative space broadly. The control, ownership, and monetization of creators’ work without their benefit is already being normalized in the discourse of generative AI companies: while some victories have been won, such as the $1.5 billion settlement between Anthropic and book authors in September 2025, closer examination reveals its limitations—the ruling simultaneously affirmed that AI training on legally acquired works constitutes fair use, many authors were excluded due to copyright registration technicalities their publishers failed to meet, and the per-work compensation remains trivial relative to the value extracted for model training. The settlement establishes that AI companies can simply purchase copies to avoid infringement claims, rather than compensating authors at rates reflecting training value (NPR 2025). The tension between platforms and so-called “content” creators echoes early digital marketplaces like Kindle Worlds and the ecosystems of Horizon Worlds (Meta), Roblox (Baszucki and Cassel), and Wattpad (Lau and Yuen): historically, such platforms promised exposure and publication while primarily exploiting user-generated content as the first product, and now training data. Yet the game’s insistence on mandatory human cooperation—its refusal to allow an AI partner—suggests that generative AI, however powerful, is ultimately only as valuable as the human experiences it provokes. The massive investments in data centers and AI companies will require justification through human interactions, not despite them. Split Fiction, then, captures the dilemma of a creative community eager to engage with generative tools but wary of losing control—yet again—to the machinery of corporate profit.