1. Introduction
[1.1] When we put out a call for papers on AI and fandom in spring 2023, generative AI was a hot topic of conversation. OpenAI's text generator GPT-4 had just been released, and text-to-image tools like Midjourney, Stable Diffusion, and DALL-E were being widely discussed (note 1). Although AI includes a wide range of technologies, some of which have been in use for a long time, we speculated that this new visibility of and access to generative AI tools would have repercussions for fandom, and we were not alone in anticipating the impacts of AI on fans (Lamerichs 2018). At that time, we had heard about AI-supported translation services for transnational fandoms (Kim 2021), fan artists using generative AI tools to generate fan art (Bai, this issue), fan fiction writers generating stories and sharing them online, fans interacting with simulations of characters or real people via the custom chatbot site Character.AI (Rosenberg 2023; Heimberger, this issue), and fans synthesizing celebrities' voices to, among other uses, narrate audiobooks and fan fiction, as well as interact with commercial synthesized versions of objects of fandom (Kang et al. 2022; Nyce 2023).
[1.2] We anticipated a large number of submissions based on the urgency of addressing this topic. And in fact, we did receive a wide range of fascinating scholarship examining many different aspects of AI and fandom, as we will describe below. However, there were not as many as we expected. Why submissions were fewer than we anticipated is in itself interesting and perhaps indicative of a wider challenge when academia turns its focus to fast-moving topics related to technological innovation. ChatGPT had 100 million active users within two months of launching, making it by some accounts the fastest-growing consumer internet application in history (Hu 2023). However, it takes time for considered scholarship to reflect on the full implications of such sweeping changes and for research to be planned and conducted and data to be analyzed. Despite this, the work included in this section is still groundbreaking and relevant; while change may come quickly to fan spaces, the importance of understanding how and why fan communities (and fan scholars) embrace—or not—new technological affordances is a constant.
[1.3] Since our call for papers closed, debates around AI in general and generative AI in particular have become even more widespread, in fandom and in wider culture. We share two illustrative events that have impacted fandom and demonstrate the potential for generative AI to be a disruptive force to fandom activities and values. Omegaverse, that specifically fannish worldbuilding trope originating in the Supernatural fandom in 2010 (Fanlore 2024), has, it seems, made its way into the data scraped by big AI companies. This came as a surprise to the creator and users of Sudowrite, a commercial tool that uses multiple language models (including OpenAI's GPT-3 in 2023), when the tool generated content that referred to specific aspects of the Omegaverse (Eveleth 2023). Fans were upset that their creative work had been ingested for use in language models and was now being sold to those who paid for the tool (whether they wanted Omegaverse references or not). A number of fans began to restrict their work on the Archive of Our Own (AO3), citing this as the reason, and the archive was prompted to make a statement about the use of AI, including the implementation of measures to stop some large-scale data scraping. They did, however, note that any data from AO3 that has already been included in existing datasets cannot be removed (https://archiveofourown.org/admin_posts/25888). For use in generative AI tools, fan labor had been co-opted and fan works removed from their intended audience without permission, and fans were not happy about this transgression.
[1.4] In our second example, in late 2024 Cliff Weitzman scraped hundreds of works of fan fiction from AO3 (as well as out-of-copyright literary works) and used AI to generate book covers, summaries, and audio versions on his for-profit website WordStream, creating outcry among fans (Minkel 2024) and prompting another round of fans to lock their fan fiction from public view. These actions by individuals and organizations outside of fandom shine a light on some of fandom's long-held values and beliefs around permission, monetization, labor, and creativity (see also Cisternino and Radillo, this issue).
[1.5] Weitzman's activities expose an additional tension: normativity. The AI-generated book covers that WordStream paired with fan fiction overwhelmingly depicted ostensibly heterosexual couples, despite many of the fan fiction works depicting queer relationships. Where fan works often try to diversify the types of characters, relationships, and stories that can be told—in opposition to the uniformity and heteronormativity of the most popular global media franchises—generative AI is designed to provide predictable responses that reflect the biases and norms of the training data used. Generative AI models are tools of normativization that reproduce the norms and representational imbalances in media (Gillespie 2024; Harding and Wagner, this issue) and exclude many populations and their perspectives (Ray Murray 2025), something that is at odds with the aims of many fan creators. The biases and normative tendencies that underpin learning models across many different types of AI are a key discussion point in wider academic discussions of these technologies (e.g., Noble 2018; Benjamin 2019) and something that we as fan studies scholars need to pay close attention to, both in terms of how fan practice can model forms of resistance and how fandom interacts with wider ecosystems around content that may come to be increasingly guided by AI and its inherent biases.
[1.6] Fan production is often spoken of in terms of transformation. The Organization for Transformative Works stresses the importance of fan works as being transformative, that is, a term related to US copyright law that allows for the use of copyrighted material if it is presented in a different manner or for a different purpose from the original and therefore transformed. Transformation implies that some labor has been performed, and affective labor performed by fans is often undervalued (De Kosnik 2009). The use of generative AI in this context, however, raises new questions about what is meant by value, transformation, creativity, and ownership. Norms of the fan gift economy hold that fan fiction, for example, is a legitimate practice in part because it is not undertaken for monetary gain. One of the ways fan creators avoid legal challenges is by not seeking financial remuneration for their work, something that is strictly enforced by AO3 (https://archiveofourown.org/tos_faq#commercial_subplatform). WordStream and Sudowrite, by soliciting subscription fees, are in direct violation of this, raising the question: If creators of fan works are not compensated financially, why should others be able to make profit from work derived from them through the use of AI? These examples are two in a long line of different attempts to monetize fan fiction (Stanfill 2025), and such practices also violate norms of permission and the attribution of sources that are key to acceptable transformative use.
[1.7] Aside from legal and financial questions around compensation and ownership, such discussions also provoke questions relating to creativity and art. The concepts of creativity and originality have been debated for millennia. Despite academic arguments made by poststructural theorists (see Allen [2011]) about the communal dimensions of authorship and the ways in which all texts rely on other texts for their meaning, the idea of creativity as original work by an individual still holds sway, in part because of the way copyright legislation operates. The creation that takes place within fan spaces—where the communal dimensions of authorship and the visibility of chains of citation and attribution are key—challenges this assumption.
[1.8] The introduction of AI also challenges this assumption if we admit that technologies play a role in creation. Celis Bueno et al. (2025) redefine creativity in the context of generative AI and argue that creativity can be seen as arising from "the interactions between technologies, practices, and social arrangements" (339). Such "a relational and distributed form of agency" (349) admits that humans, other artworks, social contexts, and technologies are necessary for creation. Li and Pang (2024) propose conceptualizing these complex relationships within participatory culture in terms of human-community-machine interactions (HCMI), which they build on in this issue.
[1.9] Technology and art have always existed together, with new technologies like photography (and later Photoshop) being hailed as the death of painting (and later photography), yet all of these forms continue to be used. However, generative AI introduces new questions around creative agency that fans are currently grappling with in terms of, for example, whether a story written by a large language model could be considered a valid form of fan fiction (see Cisternino and Radillo, this issue). Certainly, it is, as we have seen, quite possible to ask these models to produce derivative text that recognizably draws from media sources to transform them into a new text. Chiang (2024) suggests, however, that generative AI is not likely to become a new technological medium for artistic creation in the way that, say, photography is, because it does not allow for creative expression and choice-making as these other technologies do. He suggests that it is not the quality of the output that matters but the intent of the human originator to communicate—something that with AI exists in the prompt but is then filtered, mediated, and diluted by the normalization of the language models. A thousand works of fan fiction may have the same characters, setting, and basic plot, but the choices the author makes reveal something unique about their affective response to the material—something AI cannot do in its current form.
[1.10] As has happened many times in the past, fans are acting as the bellwether for societal adoptions of and challenges to new technology (Booth 2016). We are now seeing similar questions around ownership, value, and the acceptability of creative content for training AI arise in many different contexts and jurisdictions. For example, in 2024 the UK government held a consultation relating to potentially amending legal approaches to copyright and AI (Intellectual Property Office 2024). In their response (which one of the coauthors of this editorial contributed to), Imagination Lancaster, the design-led research center at Lancaster University, emphasized the importance of rights holders and called for the ability for rights holders to opt in to their work being used in AI models (Snooks et al. 2025). Many other respondents to the consultation made similar arguments, but it is yet to be seen how the government responds. In the United States, entertainment conglomerates Disney and Universal have recently sued text-to-image generator Midjourney for copyright infringement (Knibbs 2025). The verdict of this case and subsequent copyright legislation around AI could radically change which tools—and which datasets those tools are trained on—remain available for wider public use (see Mussies [2025]). We are also seeing AI put to creative fannish uses which may be counter to the desires and needs of others, causing ethical clashes between fans and for researchers (Li and Pang, this issue).
[1.11] As technology and fan studies progress, we hope also to see more fan scholars (both inside and outside the academy) explore the use of AI tools to further understand and analyze the practices of fandom, with AI as a research tool for analysis and production beyond the binary positions of technosolutionist hype or dystopian dread (Neugarten, this issue). It will also be interesting to see how fandom (and fan studies), which have often positioned themselves as aligned with social justice causes, consider wider ethical implications such as sustainability and energy uses contributing to climate change.
2. Articles
[2.1] In this special section, we see a range of approaches for engaging with the interface between fandom, fan studies, and AI, which explore but also go beyond the generative AI that has focused so much recent attention. Recurrent themes across all these articles are ethics, methodological rigor, and care.
[2.2] In the first article, Julia Neugarten has explored the potential of a suite of natural language processing and machine learning tools to examine power and gender dynamics in fan fiction featuring characters from Greek mythology. Neugarten finds value in the tools but cautions that they are most useful when used in conjunction with other methods. Using language models and computational tools requires us to ask fundamental questions about the capabilities of language, what literary writing is doing with language, and how it relates to the world. This article, by examining the potential and limitations of these tools, assesses its use in understanding fan fiction while also initiating larger debates around literary language.
[2.3] Alison Harding and Travis Wagner also look at textual dynamics in fan fiction, in this case using traditional analyses to examine work that is created using generative AI. By analyzing fan fiction output by ChatGPT, they examine how these technologies can reproduce normative ideologies relating to queerness and fail to represent the nuance, depth, and textual engagement of human-written fan fiction. A particularly interesting secondary observation in this work is that ChatGPT clearly reproduces structure and formatting indicating that fan fiction sites (specifically, AO3) are included in the training data for these models.
[2.4] Tara Heimberger moves beyond static fan fiction to examine immersive character chatbots that allow fans to carry out a conversation with fan objects. As with many of the articles in this section, she highlights ethical challenges with the use of generative AI technologies that are becoming an increasing topic of discussion in fan and academic spaces. In this article, the potential of Character.AI to disrupt the reader/writer dichotomy is explored (note 2). The author describes how fans can interact in what she calls a fully immersive way with a chatbot simulating a beloved character or celebrity. She argues that by crafting a personality to their specifications, they thereby construct a self-insert experience in real time. However, Heimberger raises questions regarding the dangers that can arise when emotional attachments form with these systems.
[2.5] In the last of the articles, Eva Cheuk-Yin Li and Ka-Wei Pang similarly highlight dangers that can arise when trust is (mis)placed in the outputs of generative systems and identify how creative fan practices can be a form of resistance against this, but with unintended consequences. They take as their case study the involvement of generative AI in producing deepfakes, specifically in the context of the Thai GL and BL celebrity video scandal. This article also takes a reflective approach to methodological questions and highlights ethical dilemmas of the research itself. By proposing a speculative ethics of care for fan studies researchers engaging with the use of generative AI by fans, they signpost new directions for fan studies research in this area.
3. Symposium
[3.1] In this special section, we also include two symposium pieces, which act as provocations toward new theoretical approaches to considering AI and fans.
[3.2] Irissa Cisternino and Rebecca Radillo's prescient piece introduces the stakes and dimensions of the debate around generative AI for fandom and the larger challenges presented by big tech. They ask us to consider major questions about how generative AI problematizes existing power structures in and around fandom, especially the importance of fan creation as a labor of love, as much as the end product of that labor. When AI bypasses that process, it breaks these community norms. Their call to action, or perhaps call to reflection, asks for increased consideration of both the media and tech industries that underpin fan activities and presents a new theoretical framework of (un)creation to understand these intersections.
[3.3] Finally, Jing Bai asks us to reconsider the relationships between humans and AI. Bai brings a personal perspective from otaku communities, describing the database model of otaku practice and how this can be mapped to new forms of human-machine interaction. Bai argues that users interacting with waifu-generating AI services are extending existing practices of information sorting and delivery rather than something entirely new. Perhaps for this reason, these users appear to extend their perception of fan practice to include AI, treating it as they would a fellow fan and otaku. This suggests that new configurations of relationships between humans and technologies are developing.
4. Conclusion
[4.1] The articles and symposium pieces in this section configure the relationship between (various implementations of) AI and fandom in multiple ways. The interrogation of AI in relation to fandom offers an opportunity to feel out the specific ways in which these new technologies, uses of technologies, and attitudes to technologies have consequences for fan spaces, fan activities, fan production, and fandom values. AI offers a site from which to examine our culturally held beliefs about what it means to be human (or nonhuman), to create, and to collaborate—topics that have long been of importance to fans and fan scholars.
5. Acknowledgments
[5.1] A huge thank you to the authors of the articles and symposium pieces in this section for working with us on this fascinating and rapidly changing area, and to the journal editors for supporting this special section.