Close Menu
    Screenwriting

    Frontiers | Narratives generated by artificial intelligence: an ethical reflection on screenwriting in contemporary cinema

    By No Comments44 Mins Read
    Frontiers | Narratives generated by artificial intelligence: an ethical reflection on screenwriting in contemporary cinema

    Narratives generated by artificial intelligence: an ethical reflection on screenwriting in contemporary cinema

    • Department of Social Science and Humanity, Miguel Hernandez University, Elche, Spain

    Abstract

    The increasing use of artificial intelligence (AI) systems in the writing of film scripts raises ethical questions relevant to contemporary cultural creation. This article analyzes these implications based on a critical review of academic literature, international regulatory frameworks, and analysis of three cases linked to generative tools applied to audiovisual narrative development: Sunspring (2016), The Diary of Sisyphus (2023), and The Last Screenwriter (2024). Using a qualitative approach, dimensions such as authorship, originality, algorithmic biases, transparency of creative processes and the work impact on professional screenwriters are studied. The results show that, while AI can optimize phases of the creative process, it also introduces significant risks related to the dilution of authorial responsibility, the reproduction of stereotypes and technological dependence in cultural industries. These issues require film-specific ethical governance, based on principles of human oversight, authorship recognition and accountability, to ensure responsible use of AI that preserves artistic integrity and the rights of creators.

    1 Introduction

    From their origins, the value of audiovisual productions lies in the originality of new ideas and new stories, even in the way they are told. The script itself is not a product for sale, it is not a product with its own wings, but it is a text that reaches its meaning once it has been shot. We start from the most basic description, the one proposed by when he stated that the script is a text that collects a story narrated in images that follow one another through a common thread. That thread is in turn formed by three others: dialogue, description and dramatic structure.

    Generative artificial intelligence (AI) consists of a set of computational techniques designed to produce original content from patterns learned in large data sets, including text, images, audio, and code (). In the field of natural language, this paradigm has been consolidated by the so-called large language models (LLMs), based on deep neural network architectures, especially transformers, which have demonstrated an unprecedented ability to model complex linguistic dependencies (). More recent research has shown the potential of these models to generate coherent texts, adapted to the context and with emergent properties, such as few-shot learning (). Likewise, foundational models constitute general cognitive infrastructures with cross-cutting applications in multiple creative and scientific domains (; ). In the field of narrative generation, these technologies make it possible to automate writing processes through probabilistic sequential prediction mechanisms, which opens up new possibilities in the creation of film scripts, from plot ideation to dialogue elaboration, while reconfiguring traditional creative production flows.

    Since its origins, cinema has progressively integrated technological advances that have transformed both industrial processes and narrative forms. The introduction of synchronised sound, colour, digital editing or computer-generated visual effects not only modified the technical conditions of production, but also influenced the cinematographic language and associated creative practices (; ). The incorporation of artificial intelligence systems is one more step within this technological evolution, now applied to early stages of the creative process, such as the writing of the script. Traditionally, the work of screenwriters has been conceptualized as an eminently authorial activity, focused on the creation of narrative structures, character design, dialogue writing, and the articulation of audiovisual action according to specific industrial and aesthetic conventions (; ). Various studies have underlined that screenwriting combines technical skills with complex cognitive processes linked to creativity, narrative empathy and the construction of audiovisual meaning (; ).

    The first applications of artificial intelligence in screenwriting were oriented towards relatively limited tasks, such as the automatic generation of dialogues or descriptions of scenes, using statistical or neural models trained on specific textual corpora (). These initial approaches mainly sought to emulate local linguistic patterns rather than global narrative structures. However, more recent research indicates that current generative systems can address more complex narrative tasks, including the creation of complete plots, character development, and the structural organization of audiovisual stories (). This change has been driven by the emergence of large-scale generative models, such as ChatGPT, capable of producing long, coherent texts that are adaptable to natural language instructions. One of the interests of academic criticism is, precisely, to analyze the incorporation of artificial intelligence systems in audiovisual writing processes, particularly with regard to narrative generation assisted by computational models. In this way, the degree of narrative complexity that these systems can reach is examined, as well as their integration into creative workflows traditionally dominated by human intervention ().

    What is evident is that AI can dramatically speed up the writing process: one study found that AI-based systems could produce an entire script in about an hour, compared to the 12 weeks that a professional writer might require (). Although these scripts tell coherent stories, AI still struggles to produce a truly original narrative: more recent research has found that writers who use AI as inspiration produce less novel and more formulaic sets of stories (). Therefore, studios have treated, until now, more as a collaborator than a replacement in screenwriting. conducted an experiment with 500 participants, presenting them with synopsis, directing and casting choices generated indistinctly by AI or humans, and found an absence of relevant perceptual bias. These findings suggest that, in terms of audience acceptance, AI-generated creativity can be perceived in a human-like way, opening up the possibility of integrating it into different stages of production without compromising audience perception.

    2 Risks and challenges of using AI in film content creation

    The ease that AI offers when creating new content expands to any discipline and highlights the need to rethink the ethical limits of this new tool. Although most citizens are still not aware of how far this tool can go, there are engineers such as Blake Lemoine, who was fired from Google for suggesting that the LaMDA language model he worked with was conscious and sensitive (Bender, 2022, cited in ). And although there is certain:

    a tendency to attribute cognitive abilities to AI that it does not have, matching them to the representation of it that is reflected in films and science fiction stories. But this trend is nothing more than an illusion of explanatory depth, a cognitive bias that has deformed the public perception of AI to the point of projecting an exaggerated, anthropomorphic and deified vision of it, on more than a few occasions (, p. 9).

    It is indisputable that humanity is being captivated by these systems that, without a doubt, are a help for many creative sectors, especially in the world of advertising and art are being widely used. Using deepfakes or the works of artists without attribution, which raise ethical debates about image rights and copyright, among others. Several studies (; ) have observed ChatGPT’s ability to create scripts similar to those that can be made by human scriptwriters and “highlight that AI can produce script elements that are often difficult to distinguish from those written by humans, and in some cases, may even be preferred by users” (, p. 30).

    But for Claudia Giannetti, AI lacks emotion, something fundamental in the connection between art itself and the audience; nor does it have autonomous thinking and therefore does not have the ability to create something original by itself (, p. 204). In this sense, human action becomes indispensable. And while AI can speed up the iteration of ideas and strengthen analytical thinking in the creation of film scripts, its use tends to privilege structural logic over the emotional engagement inherent in the creative process. The most significant ethical risks include cultural bias (predominance of Western perspectives), potential copyright infringement through model training with unpaid works, and the digital divide in access to advanced technologies (). Some studies show that some writers may be interested in AI assistance to some extent, but setting clear boundaries with AI to feel comfortable using such systems (). Experiments have even been carried out to see if personnel specialized in script creation could detect if the sample that was shown to them had been made with AI or not and the results indicate that the success rate was 52%, demonstrating that “AI was able to produce script elements that were close to the humangenerated ones, but not exactly the same. The AI may have used some patterns or formulas that were common in the script elements, but also made some mistakes or inconsistencies that the human scriptwriter would not” (, p. 18).

    The film industry has integrated AI tools into various stages of production, from rejuvenating actors to predicting commercial success. As an example, studios like Warner Bros. use systems like Cinelytic to predict box office success based on casting and synopsis variables. In the case at hand in this research, ScriptBook-derived Generative Script AI DeepStory uses a dataset of more than 30,000 scripts to act as an “artificial co-creator.” And this has led to a union response, such as that of the Writers Guild of America (WGA), which went on a five-month strike to negotiate protections against the use of AI in script development, reflecting deep concern about the impact on writers’ livelihoods (). This strike culminated in an agreement that included: the control and regulation of a technology that can already perform tasks previously reserved for authors; allow the voluntary use of AI by writers, but prohibit companies from forcing its use or AI-generated material from undermining the credits and rights of human writers; and to highlight technological duality, since writing is a technological process inserted in social and political power relations ().

    In terms of biases and cultural identity, companies such as OpenAI have recognized that models such as ChatGPT can perpetuate stereotypes and are biased towards Western perspectives, so it can be transferred to the characters and the Western perspective that would permeate a script that has been made with this technology. In addition, it perpetuates dominant narrative structures as AI tends to replicate Hollywood’s “hero’s path” structure as a universal standard, ignoring national mythologies (such as Australian) that do not always seek the hero’s mastery or self-knowledge. But also, depending on the data scraping that is done, it could make scripts with cultural appropriations of minorities or introduce decontextualized knowledge of other cultures that can damage the image of these social groups ().

    We start from the premise that a machine is more objective than a human being, but in the whole process of the AI cycle we can find biases. One of the most common is algorithmic bias, which “refers to the systematic predisposition that can arise in AI models due to biased training data or their design” (, p. 113). describes several types: data, algorithmic, user, representation, historical, sampling, aggregation labeling, confirmation, evaluation, feedback loop, demographic, institutional, and temporal. This shows how the problem is multifaceted and several can act at the same time without being exclusive (p. 15–17).

    The biases of AI are also shown from the gender variable, it is not surprising that the voices of personal assistants such as Alexa, Siri or Cortana have female voices and names, maintaining the stereotype of secretarial or personal assistance functions, which in real life are usually mostly occupied by women (, p. 314). Language is not free from the biases inherent in AI, as it is the raw material with which large language models are fed/trained. They do this through different texts on the internet, including websites, books, news, academic texts and social media content. However, in the case of Spanish, today’s language models fail to show the linguistic diversity of this language, but there is “a dialectal and sociolectal sieve [..] that includes and makes visible certain dialects while excluding and making others invisible” (Company Company, 2019, p. 109 op. cit. in .

    Legally, there is already legislation, although very general, to address this problem and counteract the possible biases that may appear with the use of AI in public administrations and companies in Spain. Article 23 of Law 15/2022, of 12 July, on the Comprehensive Law for Equal Treatment and Non-Discrimination (BOE of 13 July) establishes that:

    Public administrations will favour the implementation of mechanisms so that the algorithms involved in decision-making used in public administrations take into account criteria of minimisation of bias, transparency and accountability, whenever technically feasible. These mechanisms will include their design and training data, and address their potential discriminatory impact. To achieve this end, the carrying out of impact assessments that determine the possible discriminatory bias will be promoted.. […] ()

    Finally, it is important to highlight other challenges that should alert the screenwriter who uses AI. The most relevant is to redefine what plagiarism is, since we no longer start from absolute originality but from recognizing co-creation with the algorithm (; ; ). On the other hand, and as we have mentioned before, if we are aware of the biases and stereotypes that AI perpetuates, the screenwriter must develop a critical awareness to identify and subvert these trends. And finally, it must be aware at all times that it can be manipulated by AI itself, which in its mirage of humanization may have errors that are not detected by humans ().

    Without a doubt, the idea of authorship is rethought with these tools, but it is one of the most important challenges posed by the use of AI in any creative process. If we start from Michel Foucault’s perspective to point out that the notion of “individual author” is a historical construction that could be disappearing. According to Foucault, the author’s role regulates the flow of discourse under certain constraints but with generative AI, these constraints are shifting towards a yet-to-be-determined system ().

    Sources consulted in the scientific literature agree that AI regurgitates from what it drinks from, but currently lacks its own creative potential. The prompt can be trained to develop the story, with correct descriptions within a historical context given by the human being, contextualized by him in a logical or illogical place. AI does not understand subliminal messages, the message of looks, irony, the value of long silences. Not yet knowing how to create subtext, which as described by is where the art of good scriptwriting lies. We are aware that well used it is a tool of great help. In fact, he doesn’t hesitate when we ask him about the role of the prompt versus the script and he replies that the first is the compass and the second is the map. The answer could be described as objective, honest and accurate. The compass doesn’t take us anywhere. In short, the quality of the AI’s production depends on the quality of the prompt that the individual gives it, so human supervision is essential for the proper use of this tool.

    3 Methodology

    This study adopts a qualitative approach of an exploratory and comparative nature, aimed at the analysis of the integration of artificial intelligence systems in the writing of film scripts. The analysis of three case studies has been developed, a strategy widely used in audiovisual media research and cultural production studies to delve into emerging phenomena in specific contexts. In this case, three audiovisual productions – , The Diary of Sisyphus (2023) and The Last Screenwriter (2024) – have been selected for their relevance in the documented application of generative AI systems in screenwriting, as well as for their representative value within the different evolutionary phases of this technology.

    The selection of these three cases responds to an intentional sampling based on criteria of theoretical relevance and availability of information on production processes. In particular, variables such as the type of AI model used, the degree of autonomy in the generation of the script and the production context have been considered. This diversity allows for a comparative analysis that ranges from pioneering approaches to recent developments in the use of AI in audiovisual storytelling. The analytical process has been structured based on a textual and contextual analysis of the scripts and the resulting audiovisual works, complemented by the review of secondary academic sources and specialized criticism. On the textual level, aspects such as narrative coherence, character construction, dramatic structure and the use of language have been studied, with the aim of identifying patterns derived from the use of generative systems. On the contextual level, the conditions of production, the type of interaction between humans and AI and the critical and public reception of each work have been considered. This approach makes it possible to evaluate the technical capabilities of the models used, as well as to understand their impact on creative processes and on the perception of authorship.

    For the study of the selected works, a qualitative and prospective analysis model has been designed based on the previous bibliographic review. Given the emerging nature of AI in narrative creation, this approach does not seek to quantify data, but to identify the dynamics of co-creation, the transformations in the role of authorship and the ideological derivations of technology. Four specific axes have been selected that cover the full spectrum of the phenomenon: the technical infrastructure that makes writing possible, the reconfiguration of human work, the aesthetic-narrative result and its ethical impact.

    The rubric (Table 1) with its levels of qualitative manifestation is detailed below:

    Analysis element Level 1: Fragmented Level 2: Transitional Level 3: Integrated
    Narrative and technical autonomy Dependence on recurrent models and linear generation based on statistical proximity of words without long-term memory. Intermediate autoregressive models and hierarchical generation by layers. State-of-the-art massive language models with the ability to maintain character arcs and complex plots.
    Dramatic semantic and logical coherence Absence of causality, disjointed or fragmented dialogues. Solid structure and continuous rhythm, but with slow or mechanical passages. Fluid structural and emotional thematic coherence, capable of imitating the tropes of drama.
    Human dynamics and agency Subsequent intervention through the endowment of meaning in the staging. Directed co-creation where the human being acts as a promptor and the AI generates the bulk of the text. Curatorship and editing where the human being selects, cuts and assembles.
    Bias reproduction and amplification Automatic replication of basic genre formulas by dragging the original corpus. Eurocentric fixation on male existentialism. Crystallization of the Hollywood cliché and female domestic asymmetry.

    Qualitative analysis rubric.

    Own elaboration.

    On the other hand, a bibliographic review has been carried out on the ethical aspects of the process of creating scripts, choosing academic articles and communications to conferences of the last six years. The articles in the sample are mostly in English, where there is a little more academic production on the subject, although it is an incipient issue in research. With this review we aim to understand the state of the art with the challenges and risks presented by the use of this technology, as well as to provide suggestions to integrate them into creative processes in an ethical way.

    4 Results

    4.1 Review of audiovisual productions in which AI has been used for the creation of the script

    The analysis of specific works in which artificial intelligence has been used for screenwriting allows us to directly observe how these technologies impact narrative writing and the creative dynamics between machine and human authorship. In this way, the technical capabilities of the generative models employed and the human intervention decisions needed to transform algorithmic outputs into coherent cinematic products can be examined together. This approach offers a framework to evaluate the different methodological approaches, the levels of narrative autonomy achieved by each system and the strategies implemented to integrate AI into the creative process. Specifically, we focus on the following audiovisual productions: , The Diary of Sisyphus (2023) and The Last Screenwriter (2024), which provide a representative perspective on the evolution of AI in screenwriting, from pioneering experiments to contemporary applications with advanced language models, opening up a comparative analysis that illuminates both the possibilities and the Current limitations of these tools in the cinematographic field.

    4.1.1 Sunspring

    Sunspring represents one of the earliest milestones in the practical application of AI to film writing, being a science fiction short film designed to delve into the ability of an algorithmic system to generate an entire script without direct human intervention in the initial text. Conceived by filmmaker Oscar Sharp in collaboration with AI researcher Ross Goodwin, the project was originally developed by the Sci-Fi London 48-Hour Film Challenge, a competition in which creative teams produce short films within two days with certain requirements given by the festival. In this case, Sharp and Goodwin built a system based on a recurrent long-term memory neural network (LSTM) called Benjamin, feeding it a corpus of science fiction scripts from the 1970s and 1990s so that it would learn writing patterns and textual structure from existing examples.

    Benjamin was specifically trained to recognize and reproduce the narrative and linguistic forms present in the fed scripts, including scene prompts, dialogues, and sentence structures, allowing him to generate text in film format. After setting a minimum set of conditions—such as the Sunspring title—, a Environment Futuristic with high Rate of unemployment y a Scene initial with a Book — The Algorithm produced a Script of approximately nine minutes that included so much Dialogues as Notes of Scene. The Result It was Interpreted by Actors real, between them Thomas Middleditch, Elisabeth Gray y Humphrey Ker, bajo la Address by Sharp, Who had that interpret y Give Consistency dramatic a Quotes that many times resultaban fragmentadas o surrealistas, como órdenes de escena que desafiaban las convenciones habituales de la narrativa cinematográfica convencional.

    According to specialized journalistic analyses, the film presents a narrative that many describe as unusual or incoherent from the traditional narrative point of view, with sequences that seem to emerge more from textualized statistical patterns than from a deep semantic understanding of dramatic structure (; Bonus Stage MX, 2016). For example, contemporary critics highlight that some lines of dialogue or scene indications are strange or disconnected, which forces the performers to provide a performative interpretation that contributes to the narrative sense perceived by the audience (; ). The beginning of the short film already exposes the lack of narrative coherence: “In a future with mass unemployment, young people are forced to sell blood. That’s the first thing I can do. You should see the boys and shut up. I was the one who was going to be a hundred years old. I saw him again. The way you were sent to me.. that was a big honest idea. I’m not a bright light” (, 00:01:26).

    There is a certain thematic and causal disconnection between the sentences, as well as an abrupt displacement between subjects and actions without establishing a clear narrative context. This structure reflects the limitations of the LSTM model used, which generates text based on statistical patterns extracted from the training corpus, replicating the style and vocabulary of previous science fiction scripts without guaranteeing semantic continuity or internal logical coherence. As a result, the audience’s perception of a linear narrative or character motivations must be reconstructed through human interpretation in the staging, evidencing that AI produces formally structured textual sequences that lack autonomous narrative integrity.

    This atypical nature of Sunspring has been observed in different analyses of AI and narrative creativity, which place the short film as an example of the way in which machine learning techniques can assimilate formal patterns of the film genre without necessarily achieving a narrative coherence comparable to that of humans. While generative systems can produce structured script-like texts, the results still rely heavily on human mediation to build dramatic and emotional sense (). Sunspring is still understood in many quarters as an experiment that raises questions about the role of AI in creativity, rather than as a precedent for replacing human screenwriters (). The experience of the short film, which combines algorithmically generated text with human interpretation, constitutes a significant precedent for subsequent approaches to the use of AI in screenwriting, and serves as a basis for raising broader questions about co-creation, agency, and automation in contemporary audiovisual production ().

    The short film inherits from the science fiction films of the 80s and 90s, with which AI, Western cultural and gender bias was trained. This is reflected in the classic dynamic of two men competing in a love triangle for a woman’s attention, reproducing the stereotype of the passive female figure or “reward”. In addition, a typically Anglo-Saxon melodramatic language and confrontational structure are used, which amplifies the clichés.

    4.1.2 The Diary of Sisyphus

    The second case in which we focus our focus is The Diary of Sisyphus, an experimental feature film produced in Italy and directed by Mateusz Miroslaw Lis, which has been highlighted as the first feature film in the world written entirely by an AI. Specifically, the script was generated using GPT-NEO technology, an open-source language model inspired by GPT-3 developed by EleutherAI. Unlike productions that use AI for partial tasks – as a support in the generation of ideas or isolated dialogues – in this case most of the script was produced by GPT-NEO from instructions that included a synopsis, basic storylines and descriptions of desired scenes; the only exception was an opening monologue that was not generated by AI.

    The film’s narrative follows Adamo, a young university student in existential crisis who embarks on a journey in search of vital meaning, facing encounters and situations that emphasize humanistic and philosophical elements rather than genres traditionally associated with science fiction. This thematic approach breaks with the common expectation that AI projects applied to film are limited to genres traditionally aligned with science fiction, underscoring the possibility that generative systems can address emotive and reflective themes if provided with adequate initial guidelines.

    The film was first presented at the Festival dello Spettatore in Arezzo in November 2023 and later at the Rome Videogame Lab 2024, and had the participation of institutions such as Rai Cinema and Cinecittà. Finally, it was distributed in Italian theaters from January 2024. The use of GPT-NEO for script generation involves a process in which the AI not only generated individual dialogues, but structured entire narrative sequences based on the input prompts provided by the creators, which is a documented case of the application of generative AI in film scripts beyond laboratory experiments or short-form short films. describes the film as a work of “mechanical poetry” due to its style and rhythm and underlines the presence of philosophical influences, such as references to the thought of Albert Camus and Jean-Paul Sartre within the plot, despite the fact that certain narrative moments can be perceived as slow or surreal. The plot refers to a “sophisticated copy and paste based on an immense literary archive” that, inevitably, depends on the contributions of human beings. In the reinterpretation of the film, Sisyphus changes the rock that he is forced to drag according to Greek myth by the machine, heavy and fearsome. The intervention of artificial intelligence transforms the narrative experience by generating a character whose voice simultaneously reflects mechanicality and existential questioning. The work therefore poses a tension between textual automation and philosophical content, in which the narrative produced reflects themes of alienation, mechanicality, and the search for meaning, themes that we would undoubtedly associate with human reflection.

    The script for The Diary of Sisyphus was developed using a hierarchical expansion approach to narrative, inspired by the automatic story generation methods proposed by . Instead of generating the movie in a linear fashion, the AI first produces an expanded outline or synopsis, which is then broken down into sequences of individual scenes. By organizing the narrative into hierarchical levels, it is possible to control the expansion of the story more systematically and avoid the generation of disconnected scenes that characterize linear approaches to chain prompts. This method evidences that, although AI can produce a structurally sound script, human mediation remains crucial to ensure dramatic coherence, aesthetic intent, and narrative adequacy to the final film production ().

    The film carries a notable Eurocentric and gender cultural bias by structuring its existentialist narrative exclusively around the trope of the male hero in crisis. AI replicates the traditional view of Western literature and cinema where the intellectual journey and the search for the meaning of life belong to the male, while the few female figures are relegated to secondary roles of accompaniment or ephemeral romantic interest. On the other hand, the philosophical fixation on authors such as Camus or Sartre and the biblical elements show a training bias rooted in the Western academic canon.

    4.1.3 The last screenwriter

    Finally, The Last Screenwriter is a Swiss feature film directed by Peter Luisi and written entirely by ChatGPT 4.0, the advanced version of OpenAI’s language model, based on a series of structured prompts provided by Luisi himself. The project emerges as an experimental exploration of the creative capabilities of AI, with the stated intention of investigating the extent to which generative tools can produce a feature film script that maintains thematic, structural, and emotional coherence. The writing process began with the initial prompt: “Write a plot for a feature film in which a screenwriter realises he is inferior to artificial intelligence”, from which ChatGPT generated the plot, the characters, the narrative schemes and, finally, the scenes and dialogues that make up the complete script. Human intervention was limited to selecting and shortening scenes produced by the AI, without altering the content generated, decisions that Luisi and his team consider to be more typical of film directing or editing tasks than writing.

    The film’s narrative centers on Jack, a successful screenwriter who faces a professional and existential crisis when he discovers that an advanced automatic writing system equals or even surpasses his ability to generate stories, including emotional and empathetic elements. This plot not only constitutes the dramatic core of the film, but also replicates the very condition of production of the work, in which AI is placed as the protagonist both inside and outside the cinematographic fiction.

    The premiere of The Last Screenwriter generated public controversy and rejection in some sectors of the audience and the film industry. The scheduled screening at London’s Prince Charles Cinema was cancelled after receiving more than 200 complaints on social media and messages from viewers concerned about the use of AI in the generation of the film’s full screenplay. This highlighted the tensions around the use of AI in the creation of cultural content. Although the film was withdrawn from that specific screening, it was subsequently released for free online on July 5, 2024, accompanied by public documentation on the prompt-based writing process and the attribution of authorship to ChatGPT 4.0, which is a notable aspect from a perspective of methodological transparency in AI projects applied to cinema. In terms of production, the film was shot in January 2024, with a professional reparation from the United Kingdom and an approximate budget of $850,000, financed in part by Swiss film funds. Luisi’s decision not to allow alterations to content by cast or crew underscores an explicit commitment to the integrity of AI-generated text. The film sits at the intersection between technological experimentation and artistic discourse on creative agency, questioning the limits of human authorship and the potential roles of AI in filmwriting.

    The film structures its plot around the Hollywood cliché of the misunderstood creative genius, embodied by a middle-aged white man. The AI replicates the classic asymmetry of Western roles by relegating female characters to a purely domestic and emotional support role, stuck in the stereotype of the woman who does not understand the intellectual obsession of the protagonist.

    5 Discussion and conclusions

    In this research we have shown how the use of AI influences cinematographic audiovisual creation. With the analysis of the works described, Sunspring, The Last Screenwriter and The Diary of Sisyphus, it can be seen that the transformation affects not only the way the scripts of the works are constructed, but goes further, taking root in the work system itself, in each process, from the germ to its final development. However, it is not a parasitic relationship. It is evident that the industry has enriched itself with this system, creating intentional proposals for algorithms to work consciously in favor of the creation of scripts from intentional proposals of human beings. In this sense, the use of the new tool for the growth of creative action is positively valued.

    The transitional analysis of the three works brings to light a critical problem that goes beyond pure formal coherence: the intrinsic tendency of these models to perpetuate and amplify cultural, Eurocentric and gender biases. Far from operating as neutral tools, algorithms act as mirrors that enhance the narrative vices of the past. By drawing on a pre-existing cinematographic and historical corpus, the tools systematically reproduce the most conservative tropes of the industry, such as the objectification of women in competitive love triangles, the exclusivity of men in the existential and intellectual journey or the asymmetry of roles that relegates female characters to merely domestic functions or emotional support of the misunderstood genius.

    This recurrence shows that the narrative autonomy of AI, without corrective human intervention, naturally tends towards statistical cliché and ideological standardization. By rewarding the most repeated structures in their training data, generative models eliminate the discursive periphery and consolidate Western hegemonic narratives. Therefore, the discussion on the transition of the role of the scriptwriter cannot be limited to the technical optimization of the machine; it must demand a function of ethical curatorship and critical deconstruction on the part of the human author, whose mediation becomes indispensable when it comes to injecting dramatic meaning and actively subverting the ideological automatisms that algorithms inherit from the past.

    At this point, both as a result of the analysis of the scientific literature and of our own analysis, we are aware of the criticism aroused by the paradigm shift, from scriptor to promptor. The path of the screenwriter is intuited towards a training more aimed at the creation of orders for the system than in direct elaboration. We could see the similarities with the concept of the loss of the aura or the loss of the soul in the birth of photography. Today this thought seems naïve to us. The parallelism makes the future of the industry clear, always prioritizing economic profit, lowering costs and dispensing with human beings when it is the machine that can do the work for less expense. The transformation points to a new training of screenwriters where creativity has its origin elsewhere, but without ceasing to be the starting point. The essential thing is that the perpetrator remains the human being assisted by the machine and not the machine.

    It is precisely at this point that the ethical question arises. The intellectual property rights must continue to belong to the screenwriters, since, although they were the ones who initiated the creative process, they matured and consolidated it using the tool. In other words, considering AI as the author, when it is a machine, should be absurd and not at all coherent. There is no doubt that the creative thinking mind is human. Social and cultural policies must safeguard this right.

    However, the legislation in this regard is very general and is usually based on the classic principles of copyright such as the Spanish Intellectual Property Law, where the essential requirements for the protection of a work are: “that the work is the result of a creative act of a natural person, that is, human authorship and, on the other hand, that this result is original, in accordance with the provisions of articles 5 and 10” (, pp. 46–47), In general, this legislation occurs in other European countries despite the fact that these tools represent a change in the process of human creation that:

    is complicated when we talk about the use of protected data for training, since many systems, such as GPT, DALL· E or Sora, they train with enormous volumes of information, much of which includes copyrighted material: scripts, images, films, music. The legality of this depends on each country, in the US, for example, “fair use” is used to justify training, but this defense is not universal (, p.49).

    On the other hand, given the analysis, it is worrying that the absence of a motivation or standardization of the use of AI without surveillance or control may lead to the homogenization of the culture already exploited: the format of the male hero, the villain, also a man, the collaborating character generally female, heterosexual relationships, the stories always told from the North American approach as an indisputable centripetal ideology, etc., further favor cultural colonialism that invades like an invisible shadow with discourses that leave out minorities of any typology. It is evident that with the use of AI without self-criticism, narrative structures can lead to enhancing this process. Hence the importance of the screenwriting profession having a critical awareness with the ability to detect these patterns in algorithms and correct them if necessary.

    Biases can appear from the selection of data that are not representative of reality, called sampling bias, this occurs a lot from the gender perspective when AI, if the gender is not specified, speaks or creates images to a greater extent of men than women, while women are 50% of the world’s population. On the other hand, biases can occur by replicating prejudices that already exist in society and that the algorithm learns with training information (, p. 314).

    Likewise, the film industry and the societies in which it is developed must protect screenwriters in the process and be realistic about the limitations of technology. At this point, we rescue as evidence the conflict caused by the strike of the Writers Guild of America (WGA). The agreement that was reached left the screenwriters free to use AI or not. The risk, however, does not lie in the use or not, nor in whether AI will replace the human being, which is also evident will not be the case. The real risk is to consider the potentiality of technology above the human being, as if being a creator of promptors required less intellectual effort, denying or devaluing the critical function. The consequence of this risk would be the devaluation of the profession of screenwriter to a mechanical job, of a lower category and, therefore, unfairly justify a lower salary.

    It is necessary to work within an international regulatory framework, designed from the industrial, creative, legal, social and cultural multidisciplinarity. For ethical responsibility, the use of transparency techniques is suggested that evidence human and technological creation in the scripts, always being attributed to the human author since the final product is his responsibility. In other words, the attribution and remuneration must be complete and qualified. On the other hand, new screenwriters must be trained by taking advantage of the advantages of AI, mastering prompting and critical vision, a solid and indisputable cultural base.

    To mitigate biases, it is necessary to “consider transparency, interdiscipline and ethics in development teams, as well as the implementation of rigorous ethical audits based on regulations and ethical principles.” (, p. 1311–1312) Transparency is achieved by clearly showing which datasets are used to train AI, and AI companies must be accountable, which implies that:

    assume the consequences of their actions and decisions, and that they may be subject to sanctions or reparations in case they cause damage or violate rights. It manifests itself by reporting on algorithmic performance in specific subgroups and by striving to mitigate disparities (, p. 316)

    Finally, as conclusive contributions of our study, we consider it necessary and pertinent to propose a series of recommendations for the use of AI in the creation of film scripts:

    • Change of concept: from scriptor to being a promptor, which is the next step in the transformation of the author’s role in the era of Large Language Models (Gürsoy & Şavk (2025).

    • There is no co-authorship: since the final responsibility lies with the creative author. However, rein the creation of a script

    • Promotion of critical skills in screenwriters. Therefore, from their training, the skills of future screenwriters to interact responsibly with AI must be improved, understanding its strengths and weaknesses.

    • Design training curricula that include ethics with the use of AI and extend it to society in general with digital literacy and end with re-skilling and up-skilling initiatives to transition the workforce to sectors such as cybersecurity and data science that define new economic systems.

    • Work under international regulatory protection, designed from industrial, creative, legal, social and cultural multidisciplinarity.

    • Creation of a public manual, preferably under the auspices of major global institutions, that describes mechanisms for attribution of authorship, such as digital watermarks or leaving a trace in metadata, as well as the legal regulation of users and content moderation rules in general, in short, creating detectors for the use of AI ().

    We are aware that, at the time of writing this article, the specific literature on ethics in screenwriting is limited, as well as audiovisual productions that have explicitly stated their use and that have served as a case study and, contrary to the fact that it can be considered as a limitation of the study, is presented as a pioneering, prescriptive and necessary research, to lay the foundations for future work on the theme that unites the creation of scripts made with AI and ethical responsibility.

    Individual responsibility is fundamental in these processes taking into account two keys: The need for human supervision when interacting with generative AI tools and the principle of “junk in, junk out” because the results of AI are only as good as the commands (prompts) that the individual generates, so the quality of the final result is directly related to the responsibility and skill of the user. In conclusion, although AI is a collaborative tool, the weight of critical judgment and ethical direction falls on the individual ().

    Statements

    Data availability statement

    The original contributions presented in the study are included in the article/Supplementary Material, further inquiries can be directed to the corresponding author.

    MJ-M: Funding acquisition, Writing – original draft, Writing – review & editing, Conceptualization. CML-R: Investigation, Writing – original draft, Writing – review & editing, Supervision. MS: Investigation, Methodology, Writing – original draft, Writing – review & editing.

    Funding

    The author(s) declared that financial support was received for this work and/or its publication. This research and publication has been funded by the Chair of Audiovisual Analysis and Foresight of the Valencian Audiovisual Council and the Miguel Hernández University.

    Acknowledgments

    This is a short text to acknowledge the contributions of specific colleagues, institutions, or agencies that aided the efforts of the authors.

    Conflict of interest

    The author(s) declared that this work was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

    Generative AI statement

    The author(s) declared that generative AI was used in the creation of this manuscript. Analysis of bibliographic

    Any alternative text (alt text) provided alongside figures in this article has been generated by Frontiers with the support of artificial intelligence and reasonable efforts have been made to ensure accuracy, including review by the authors wherever possible. If you identify any issues, please contact us.

    Publisher’s note

    All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.

    References

    • Al-kfairyM.MustafaD.KshetriN.InsiewM.AlfandiO. (2024). Ethical challenges and solutions of generative AI: an interdisciplinary perspective. Informatics11 (3), 58. 10.3390/informatics11030058

    • AntonioE. (2016). Sunspring, the film script written by an artificial intelligence, Bonus Stage Mx, June 17. Available online at:https://bonusstage.mx/sunspring-el-guion-cinematografico-escrito-por-una-inteligencia-artificial/ (Accessed January 22, 2026).

    • BattyC. (2014). Screenplays: How Texts Shape Meaning in Film. London: Palgrave Macmillan.

    • BehroozM.TianY.NganW.YungsterY.WongJ.ZaxD. (2024). “Holding the line: a study of writers’ attitudes on co-creativity with AI,” in GenAICHI: CHI 2024 Workshop on Generative AI and HCI 3, Estados Unidos, 1–7. 10.48550/arXiv.2404.13165

    • BodenM. A. (2016). AI: Its Nature and Future. Oxford: Oxford University Press.

    • BommasaniR.HudsonD. A.AdeliE.AltmanR.AroraS.ArxS.et al (2021). On the Opportunities and Risks of Foundation Models, Stanford CRFM Report. 10.48550/arXiv.2108.07258

    • BrownT. B.MannB.RyderN.SubbiathM.KaplanJ.DhariwalP.et al (2020). Language Models are Few-Shot Learners, NeurIPS. 10.48550/arXiv.2005.14165

    • CakeS. (2025). Artificial intelligence as a collaborative tool for script development. Media Pract. Educ.26 (3), 302–317. 10.1080/25741136.2025.2454074

    • ÇelikK. (2024). AI Vs. Human in screenwriting: is AI the future screenwriter?Sakarya Commun.4 (1), 1–22. Available online at:https://izlik.org/JA44US43GU(Accessed May 1, 2026).

    • ChuE.DunnJ.RoyD.SandsG.StevensR. (2017). AI in Storytelling: Machines as Cocreators. New York: McKinsey & Company Media & Entertainment.

    • CristantoL. V.ImanjayaE. (2025). AI And scriptwriting: a new threat?J. Theoret. Appl. Inf. Technol.103 (13), 4912–4920. 10.5281/zenodo.18092337

    • CrosbyP.McKenzieJ. (2025). The reel deal? An experimental analysis of perception bias and AI film pitches. J. Cult. Econ.49, 281–300. 10.1007/s10824-025-09534-4

    • DayoF.MemonA. A.DharejoN. (2023). Scriptwriting in the age of AI: revolutionizing storytelling with artificial intelligence. J. Media Commun.4 (1), 24–38. 10.46745/ilma.jmc.2023.04.01.02

    • DoshiA.HauserO. P. (2024). Generative AI enhances individual creativity but reduces the collective diversity of novel content. Sci. Adv.10 (28), 1–9. 10.1126/sciadv.adn5290

    • EnFilme. (2016). Video: “Sunspring”, short film written by an Artificial Intelligence system, EnFilme, June 14. Available online at:https://enfilme.com/notas-del-dia/video-sunspring-cortometraje-escrito-por-un-sistema-de-inteligencia-artificial/ (Accessed January 22, 2026).

    • FanA.LewisM.DauphinY. (2018). “Hierarchical neural story generation,” in Proceedings of the 56th Annual Meeting of the Association for Computational Linguistics, Vol. 1. Association for Computational Linguistics, 889–898.

    • FieldS. (2005). Screenplay: The Foundations of Screenwriting. Delta.

    • FranganilloJ. (2022). Artificial intelligence-generated content: opportunities and threats, ThinkEPI Yearbook, 16. 10.3145/thinkepi.2022.e16a24

    • GürsoyA. Ö.ŞavkS. (2024). From scriptor to promptor: an evaluation of the status of authorship, authenticity, and creativity in light of the use of artificial intelligence in screenwriting. ARTS12, 57–82. 10.46372/arts.1482636

    • GrabeI.DuqueM. G.ZhuJ. (2022). “Towards a framework for human-AI interaction patterns in co-creative GAN applications,” in Proceeding of the 3rd Workshop on Human-AI Co-Creation with Generative Models (HAI-GEN ‘22) at ACM IUI Workshops. Workshop on Human-AI Co-Creation with Generative Models, Virtual, Finland, 21/03/2022. Available online at:https://pure.itu.dk/en/publications/towards-a-framework-for-human-ai-interaction-patterns-in-co-creat-2/(Accessed May 1, 2026).

    • KantosaloA.Thattai RavikumarP.GraceK.TakalaT. (2020). “Modalities, styles and strategies: an interaction framework for human–computer co-creativity,” in Proceedings of the Eleventh International Conference on Computational Creativity. Association for Computational Creativity, 57-64, International Conference on Computational Creativity, Coimbra, Portugal, eds. CardosoA.MachadoP.VealeP.CunhaJ. M., 07/09/2020. Available online at:http://computationalcreativity.net/iccc20/papers/ICCC20_Proceedings.pdf(Accessed May 15, 2026).

    • IslamG.GreenwoodM. (2024). Generative artificial intelligence as hypercommons: ethics of authorship and ownership. J. Bus. Ethics.192 (4), 659–663. 10.1007/s10551-024-05741-9

    • LuchenF.ZhongweiL. (2023). ChatGPT begins: a reflection on the involvement of AI in the creation of film and television scripts. Front. Art Res.5 (17), 1–6. 10.25236/FAR.2023.051701

    • MacdonaldI. W. (2013). Screenwriting Poetics and the Screen Idea. London: Palgrave.

    • ManovichL. (2001). The Language of New Media. Cambridge: MIT Press.

    • MartínezF. L.PeñaJ. H. G. (2024). AI And biases: an alternative vision expressed from ethics and law. Inform. Derecho15, 109–121. Available online at:https://revistas.fcu.edu.uy/index.php/informaticayderecho/article/view/4738(Accessed June 22, 2026).

    • McKeeR. (1997). Story: Substance, Structure, Style and the Principles of Screenwriting. New York: Regan Books.

    • Muñoz BasolsJ.Palomares MarínM. D. M.Moreno FernándezF. (2024). Digital Language Bias (DSS) in Artificial Intelligence: Implications for Massive Spanish Language Models. Available online at:https://idus.us.es/bitstreams/7d2f9f98-96d9-41c1-8c16-e0d8405cf055/download(Accessed June 22, 2026).

    • Muñoz VelaJ. M. (2024). Generative artificial intelligence. Challenges for intellectual property. UNED Law Rev.33, 17–75. 10.5944/rduned.33.2024.41924

    • NajiK. (2024). A systematic review of the digital transformation of the building construction industry. IEEE Access12, 31461–31487.

    • Pérez-Ugena CorominaM. (2024). Gender bias (in AI). Eunomia. J. Cult. Legal.26, 311–330. 10.20318/eunomia.2024.8515

    • PicoD. (2025). Authors or algorithms? The debate on creativity and intellectual property in the age of AI. INMÓVIL11 (1), 6–6. Available online at:https://www.inmovil.org/index.php/inmovil/article/view/129(Accessed June 22, 2026).

    • PrinceS. (2012). Digital Visual Effects in Cinema. New Jersey: Rutgers University Press.

    • RezwanaJ.MaherM. L. (2022) “Designing creative AI partners with COFI: a framework for modeling interaction in human-AI co-creative systems,” in ACM Transactions on Computer-Human Interaction. Proceedings of the 12th International Conference on Computational Creativity (ICCC ‘21), ISBN: 978-989-54160-3-5. Available online at:https://arxiv.org/abs/2204.07666(Accessed March 21, 2026).

    • Robina DuhartP. D. (2025). Biases in AI? Analysis of algorithmic bias and a proposed solution. Med. Ethics36 (4), 1293–1336. 10.36105/mye.2025v36n4.02

    • RuizI. (2025). “AI Is objective. An analysis of bias in AI systems,” in Myths and Representations of Artificial Intelligence, eds. BecerraG.MezzadraJ.Mo 14. ISBN: 978-987-710-129-4, 10.63790/978-987-710-129-4

    • SegerL. (2010). Making a Good Script Great. 3.ª ed. Los Ángeles: Silman-James Press

    • SunP. (2024). “A study of artificial intelligence in the production of film,” in SHS Web of Conferences, 183. 10.1051/shsconf/202418303004

    • Sunspring (2016). Directed by Sharp, O. [Film]. United States: End Cue Production.

    • VaswaniA.ShazeerN.ParmarN.UszkoreitJ.JonesL.GómezA.et al (2017). “Attention is all you need,” in Advances in Neural Information Processing Systems 30: Annual Conference on Neural Information Processing Systems 2017, Long Beach, 4-9 de diciembre de 2017, 5998–6008.

    • VV.AA. (2023). Round table on ethics and authorship in media and cinematographic creation in times of AI. Comp. Cinema11 (21), 199–214. 10.31009/cc.2023.v11.i21.11

    • VignoliV. (2023). Il diario di Sisifo, by Mateusz Miroslaw Lis, Sentieri Selvaggi, 9 November. Available online at:https://www.sentieriselvaggi.it/il-diario-di-sisifo-di-mateusz-miroslaw-lis/ (Accessed January 23, 2026).

    • Zaldívar-ColadoA.Tripp-BarbaC.Aguilar-CalderónJ. A. (2025). Artificial intelligence and cinematic creativity: review of the impact of ChatGPT on the evolution of screenwriting. Rev. Ibér. Sist. Tecnol. Inf.58, 20–35. 10.17013/risti.58.20-35

    • Zamora ManzanoJ. L.Ortega GonzálezT. Y. (2024). Ethics, law and technology: exploring the representation of artificial intelligence in cinema. Gen. J. Law Lit. Cinematogr. 1, 435–452. Available online at: https://accedacris.ulpgc.es/bitstream/10553/129169/1/%C3%89tica%2C%20Derecho%20y%20Tecnolog%C3%ADa%20Explorando%20la%20representaci%C3%B3n%20de%20la%20Inteligencia%20Artificial%20en%20el%20Cine.pdf(Accessed June 22, 2026).

    • ZhaoP.ZhangH.YuQ.WangZ.GengY.FuF.et al (2024). Retrieval-augmented generation for AI-generated content: a survey. Data Sci. Eng. 11, 1–29. 10.1007/s41019-025-00335-5

    Summary

    artificial intelligence, authorship, creative industries, ethics, film script

    Jurado-Martín M, Lopez-Rico CM and Samper Cerdán M (2026) Narratives generated by artificial intelligence: an ethical reflection on screenwriting in contemporary cinema. Front. Commun. 11:1884330. doi: 10.3389/fcomm.2026.1884330

    Katja Schmid, Stuttgart Media University, Germany

    Jorge Luis Morton, Autonomous Metropolitan University, Mexico

    Samah Nassar, British University in Egypt, Egypt

    This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) and the copyright owner(s) are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.

    All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article or claim that may be made by its manufacturer is not guaranteed or endorsed by the publisher.

    Share.

    Related Posts

    Add A Comment
    Leave A Reply