1. Introduction
[1.1] Artificial intelligence (AI) is increasingly entering the conversation about creative industries, from video games aiming to integrate AI into more of the creative process (Mulligan 2024) to concerns about the safety of creative jobs in various media industries (Reynolds 2023). Less explored, however, is the increasing role AI may be playing in fan spaces. There exists an entire fan economy of production, from video game modifications that expand player experience (Postigo 2007) to fan art and fan fiction that can reshape media and popular culture content in a variety of ways (Dym et al. 2018). AI has begun to influence and shape these areas of fan experience and engagement as well, perhaps unsurprisingly.
[1.2] While we may be familiar with the AI we encounter every day, such as personal voice assistants, the introduction of AI chatbots has offered a unique social outlet for people (De Freitas et al. 2024). These chatbots can have real time conversations with people, responding as if they were an actual person on the other end. This study is interested in newer applications of these uses of AI, specifically the ways that chatbot apps allow fans to chat with their favorite characters and create stories. This study is an investigation into online dynamics between fans in multiple video game fandoms, among users of new AI chatbot technologies, and into fans' discontents as new technologies emerge and develop.
[1.3] Given AI's presence and influence, how is AI impacting fan communities? How do AI programs shape fan behaviors, interactions, and consumption of media and popular culture? This qualitative study uses online forum discussions among users of two AI chatbot programs and drawn from online forums dedicated to three video games to explore how fans are engaging with, exploring, and using AI as part of the fan-creation process.
[1.4] Specifically, I find that these fans and users of new technologies are using online forums as a social space to discuss the new role that AI is playing in their fandom experience and reacting to the introduction of AI into their fandoms in uneven ways. AI is increasingly becoming part of the wider experience that fan fiction has offered fans, allowing for expanded narrative exploration, self-expression, and the fan experience. At the same time, however, these new technologies are not without their potential negative impacts, including concerns about the environment in addition to worries about AI training based on original fan works.
2. Artificial intelligence and chatbots
[2.1] Much of the artificial intelligence that we encounter in our everyday lives, like virtual assistants, is reliant on machine learning (Riedl 2019). Machine learning, put simply, allows programs to adapt and be trained on large amounts of data to become responsive to human needs. Increasingly, AI applications involve machine learning that also produces something new, relying on natural language processing (NLP) to understand language input and generative AI techniques to produce images based on training (Bansal et al. 2024). Some of these interactive forms of AI, like chatbots, have moved toward large language models (LLMs) to produce more accurate conversation (Naik et al. 2024) by allowing training on huge amounts of data (e.g., text and information; Naveed et al. 2023). Despite being powerful, many of these models—whether related to image generation or text generation—have also produced a number of cultural biases related to social statuses like race and gender (Naik and Nushi 2023).
[2.2] Because these models draw from bodies of data produced by people with their own cultural biases, we often see stereotypes being reproduced in information and images supplied by AI. For example, overrepresenting white people and men and a particular skew toward showing these kinds of people in positions of power or authority with fewer people of color represented and women tending to be sexualized (Gengler 2024). These biases appear to be reproduced in AI used by fans, steering conversations toward the heteronormativity and gendered reactions noted in the analysis below.
[2.3] These new technologies have led to the rise of more casual usage. Chatbots have increasingly been used for their ability to assist in information acquisition, entertainment, socialization, and just to engage with a new technology (Brandtzaeg and Følstad 2017). Chatbots also rely on machine learning (Suta et al. 2020; Shumanov and Johnson 2021) to produce what feels like a personality that users can have a conversation with (Shumanov and Johnson 2021). While many uses of chatbots have been studied for their applications in customer service (Shumanov and Johnson 2021), they have more recently increasingly been used in personal ways and have been found to alleviate loneliness for users who use them to have the sense of someone to converse with (De Freitas et al. 2024). While these technologies may initially seem unrelated to fandom and fan-created works, they are presenting new opportunities—and potential obstacles—for fandom communities. This project explores how the new plethora of personal chatbot programs is increasingly used by fans to create personalized stories that function as interactive fan fiction.
3. Fandom, fan fiction, and emotional connection to characters
[3.1] There is labor involved in fandom (Stanfill and Condis 2014), whether this is related to making resources for other fans, writing stories based on source material (e.g., fan fiction), or producing art or other pieces of media that build upon the popular culture they consume. Fan fiction in particular has been of interest in terms of this labor process and for its transformative potential, allowing underrepresented audiences to shift narratives toward more diverse representation, challenging normativity and narrow storytelling structures (Jenkins 2012) and allowing audiences to push back against restrictive messages in media (Wild 2018). This is useful for fans of video games, where representation is improving, but continues to be limited when it comes to gender and LGBTQIA+ inclusion (Greer 2013; Utsch et al. 2017). Involvement with this element of creative fandom is also highly emotional, including the process of reading fan fiction (Samutina 2017).
[3.2] Much of the drive to create and read these works is related not only to love of specific stories but of specific characters. In many ways, these practices are related to and driven by parasocial relationships, or one-sided bonds that fans feel toward characters that they are particularly connected or attached to (Nguyen et al. 2023). These elements of connection can become even more personal through fan fiction, where producers and consumers want to feel as close as possible to the characters. This can take the form of self-insert stories, where the writer imagines themselves in the story, or y/n or xReader stories that allow readers to search out stories that remove identifiable information of the main character for them to insert themselves as the protagonist (Sapuridis and Alberto 2022).
[3.3] There are limitations to the scope of creativity and agency within these fan-created works, however. While creators of fan-made works that reimagine characters and stories have creative control over the direction the stories take, readers of these stories do not have much agency beyond using keywords or searches for the types of stories they want to read, relationships they want to read about, or the media they want to see represented in these works, although there are often limitations to these functions as well (Johnson 2014).
4. Agency, input, and video games
[4.1] The interactivity of and agency involved with video games (Muriel and Crawford 2020) can enhance the closeness that players feel with characters. Players of video games are often responsible for many in-game decisions, within the bounds of a game's design (Bizzocchi and Tanenbaum 2012; Muriel and Crawford 2020). They are often able to exert some influence on a game's environment, a game's world state, and the characters within that game world, including their own (Murzyn and Valgaeren 2016; Brierley-Beare 2023). In some cases, they may also be able to forge bonds with in-game characters, which are sometimes romantic in nature (Brierley-Beare 2023; Tomlinson 2021).
[4.2] As with reading fan fiction, many of these instances of agency can also be emotional for players. The connections and bonds they form with other characters, whether they are romantic or platonic, can help steer the decisions that they make in the game (Tomlinson 2021). Many players find themselves wanting to make the best decisions for their own character (Tomlinson 2021), but also for the characters they encounter and the game world more broadly (Murzyn and Valgaeren 2016). These emotional influences on agency help players, as they provide input for the narratives they explore in video games (Moser and Fang 2015). It is important to better understand the impacts of the new levels of agency being granted to fans by new technologies like AI.
5. Methods
[5.1] This qualitative study emerged from other studies on video game player behaviors and online discussions focusing on multiple video games and game fandoms. Online discussions are drawn from official game forums as well as app- and game-focused spaces on Reddit (known as subreddits). These studies use a grounded theory approach (Charmaz 2006), allowing for data collection and observation to drive the analysis forward, rather than focusing on specific hypotheses or questions. As a result, AI entered conversations in these studies, presenting new paths for inquiry that are partly based on analysis of online conversations among players of three video games in the original sample. Online conversations among AI app users were added as part of this analysis after the relevance of AI chat programs became clear. This study is also partly theoretical in nature, conceptualizing the ways that trends in these AI apps are influencing consumption within gaming fandoms and exploring the growing impact and implications of AI technologies on fandom communities.
[5.2] Two AI programs allowing users to create their own stories are part of this study: TalkieAI and CharacterAI. Each of these programs allows users to create characters that other users can have text conversations with, in addition to generating images to go along with these characters. While CharacterAI specifically bills itself as a space to chat with users' favorite characters, TalkieAI is being used as a space to do so by users as well. While there are significant concerns about AI's impact on the environment due to the resources taken to generate content (Iqbal et al. 2024), some use of these programs is included for the study for context and understanding of user and fan experience but was limited to approximately ten hours of use across the programs.
[5.3] The three games/fandoms in this study are: Dragon Age (DA), Baldur's Gate 3 (BG3), and Love and Deepspace. These games are being studied in other projects, in part, for their inclusion of romantic content. Dragon Age (BioWare 2009–2024) is a role-playing video game (RPG) series with four major game releases, with the earliest title released in 2009 and the most recent title released in 2024. The inclusion of DA is based on twenty-five months of forum observation. It is also based on approximately 682 hours of gameplay across the series.
[5.4] Baldur's Gate 3 (Larian Studios 2023) is a fantasy RPG based on the tabletop game Dungeons & Dragons. The inclusion of BG3 is based on fourteen months of forum observation. It is also based on approximately 217 hours of gameplay for context. Love and Deepspace (INFOLD 2024) is a romantic mobile otome game (a romance game with a female protagonist) with a large global fan base. Online forum observations are based on approximately nine months of observation of online forums and gameplay is based on approximately 633 hours.
[5.5] The analysis of this data is based on an open-coding scheme (Corbin and Strauss 1990) that allowed for the production of themes from online conversations among fans and program users both during and after data collection. This resulted in the themes discussed below, highlighting some of the capabilities and potentials of AI as it becomes a bigger influence in everyday life and encroaches on more of fandom life. Although the online forums used in this project are publicly available, user privacy is protected by obscuring the exact names of the forums, removing usernames, and slightly rewording quotes from users to reduce the likelihood of identification while maintaining their original meanings and sentiments.
6. Overview of the programs
[6.1] Both programs studied in this project work under similar principles but are perceived differently by users. Users of TalkieAI do not tend to perceive this as a fan fiction generator, despite the availability of many well-known characters, including video game characters. Users of CharacterAI, on the other hand, do view this as a specifically fan fiction–like experience, allowing them to explore and expand upon some of their favorite stories, sometimes specifically noting these as self-insert stories that they can steer on their own.
[6.2] In each case, there are two elements to these programs: creation and conversation. The first element, creation, has a few factors involved. Part of this relies on users (whom I will refer to as creator-users) developing the character, while the other part depends on a large language model that is part of AI machine learning. Creator-users can create original characters or develop characters based on existing source material. This means that users of these programs are often specifically making familiar characters. This allows users to engage in full conversations with those characters. This brings the relationships they hope to have with these characters—and an extension of the fan fiction experience I am theorizing—to a new level.
[6.3] In terms of the LLM involved with this process, chatbots used in these programs rely on training, like any other LLM. To be able to have what seems like a conversation with a person, rather than a machine, the model needs to learn a number of things: how to parse language inputs, how to provide an appropriate output, and how to understand their own contexts as a character within an existing video game story and fictional universe. This contextual learning may involve scraping the internet for information from wikis, game information, and—as many users point out below—potentially even fan-created works.
[6.4] Herein lies one of the potential issues with the application of AI: Given that these programs scrape from web content based on the directions given to them to form a particular character, it is possible that some original fan works get caught amid the attempts to make these characters more believable. This could mean—and fan fiction writers are concerned that—original works are being used to inform the interactive stories that AI users create and consume without creators' knowledge or consent and without giving them proper credit. This has been highlighted by artists and AI critics as a major problem with generative AI used to create art—because the models need to learn from existing materials, they are effectively stealing content from original creators (Goetze 2024). In the case of these chatbots, fans can experiment with stories and create new dynamics with characters, but potentially at the cost of exploiting human creators.
7. AI chatbots as dynamic, cocreative fan fiction
[7.1] Despite the clear negative potential impacts, these programs are attractive to users because they present unique opportunities to not only develop original stories and characters—albeit steered and fleshed out by both AI and the users who interact with the chatbots—but to interact with their favorite existing characters. They can have active and evolving conversations that function as dynamic fan fiction with these characters, exploring stories initially crafted by creators, but further shaped and defined by the reader. In the words of one user of a CharacterAI forum, the program is essentially "interactive Wattpad," a platform used for fan fiction. Specifically, some users point to the experience as being like an interactive y/n, or your name, story.
[7.2] There is a cocreative experience that these chatbots allow as part of the dynamic nature of these stories. While someone has produced the characters and the characters are drawing from information online and from the input provided by the initial creator-user (those producing the AI characters), reader-users (those interacting with these characters) steer the story and provide their own additional input. This can lead to reader-users discovering holes in these stories that lead to dissatisfaction, which will be discussed further below, but this also provides latitude for experiences that traditional fan fiction does not. Reader-users can explore new stories, new angles on their favorite media, and relationships with characters in ways that are similar to fan fiction but have much more room for the person experiencing the story to impart their own creativity. This extends the agency (Bizzocchi and Tanenbaum 2012; Muriel and Crawford 2020) we associate with video game play to the fan fiction reading experience, where fans of video games and beyond can shape and impart their own desires and influence into the narrative.
[7.3] These programs and technologies offer fans a perfect opportunity to expand on stories or even to have a deeper, real-time self-insert experience with a character of their choosing, whether this takes a romantic turn or is simply to experience an adventure with a character they admire. Users enjoy sharing their stories with one another in both TalkieAI and CharacterAI forums, pointing out funny interactions, sharing experiences, or highlighting their complaints. They also share their own uses of these programs and compare their approaches. As one user brings up in conversation online, "I ship myself with a video game character, trying to make them as accurate as possible, so creating my own fanfic. It's the best feeling when what they say feels canon. Is that common?" (user, CharacterAI forum).
[7.4] While the use of these programs is effectively the same—an interactive story, often with familiar characters from well-known media—CharacterAI users are much more inclined to see their engagement with the program specifically as fan fiction and to use fandom or fan fiction terms such as self-insert and shipping to explain their experience. Users of TalkieAI, on the other hand, discuss the content directly and without fandom-related flourishes. They discuss prompts, responses, and characters, but do not frame these as fan fiction stories, nor do they compare their usage with spaces related to fan fiction like users of CharacterAI. It is unclear why users of these programs engage with them so differently in everyday conversation yet use them in similar ways. Regardless of what prompts users to think of these apps differently, fan fiction behaviors exist across these apps, while fan fiction language only regularly appears among users of one.
[7.5] Some of these conversations directly illuminate the differences for these users between fan fiction and even live role-play with others as well. In one conversation between users, which also highlights some of the criticisms of AI and its impact on the environment, they discuss the drawbacks of other methods of engaging with creative stories within existing narratives. A conversation on the CharacterAI forum mentions:
[7.6]User 1: I don't like when people say to just role-play or read fanfiction because I feel comfortable talking to these characters instead of trying to do this with real people. You're not going to be judged by bots for what you say, and you can put your soul into it…And I read fanfiction but it's not a substitute. And they say the bots are bad for the environment, but you have to blame the companies for that.
User 2: Fanfiction is like passive role-play. The story is written and you read it…Chatbots are more active, so you're in the story and it's always changing. If you don't know what to say, the character can take over (although it's not always the best at it).
[7.7] Users perceive real advantages to the capabilities of AI for interactive fan fiction: the possibility of role-play without judgement (or scheduling issues, as some users point out when it comes to role-playing with others) and creating their own stories with a reactive character. AI allows for the same kinds of creativity that fan fiction typically does, giving fans the ability to reimagine stories, develop new possibilities for characters, and provide opportunities for producing new elements of diversity within story worlds. There are caveats on this last element of potential that will be discussed in more detail below.
[7.8] Creator-users and reader-users are not the only ones steering these stories, however. Often, characters in each app develop their own lore, responses, and even recall elements of the stories they draw from. In my experiments with characters generated by users of TalkieAI from the video game series Dragon Age, for example, the character Cullen recalls missions with another character, Cassandra, that did not happen in the game but drew on canon lore and seemed plausible. In other cases, however, characters diverge too far from canon lore for users to become fully invested, reducing the effectiveness of AI in fulfilling the role that fan fiction can in terms of expanding narratives.
[7.9] In an early conversation among users online about CharacterAI in a Dragon Age forum, fans discuss a major character from the two most recent game titles in the series. The users express some disappointment in the program's ability to generate a true-to-character conversation:
[7.10] User 1: I made Solas in character.ai and he talked about [his in-game mission of] taking down the [barrier between worlds] right away! It's amazing. I want to see what other people experience with it. Just search for him or you could even make your own. I really want to see how well AI can imitate our beloved characters.
User 2: I started by asking if he believes in the maker. He said yes and he said that his faith keeps him going when things feel darkest. Fake.
User 1: [laughing] That's hilarious. I didn't ask about that, but he made some comments about [the elven gods] that sounded close to the game. Oh well.
[7.11] As some context, Solas is an elf and, within the game, would be more inclined to believe in the elven pantheon of gods (which he is secretly part of—a fact that is revealed later in the game series). Accuracy for players seems to be hit and miss, but when accuracy of these characters' responses falters, it draws users out of these stories quickly.
[7.12] There are many moving parts involved with these AI chatbots, and users play an active role in much of this process. They are responsible for feeding information into these chatbots to develop their personality and at least some of the characters' knowledge about their histories, lore, and source material. In the end, these users are provided with a unique fan fiction experience that offers more control over these stories as readers, producing an experience that is like a mashup of fan fiction and role-play.
8. Consuming, engaging with, and emotionally connecting to AI fan fiction
[8.1] Beyond extending stories, these chatbots also serve another function. Many users note that because of the dynamic nature of these conversations, not only do they allow for more creative expressions and explorations, but they also alleviate loneliness. Many users report talking to the characters they interact with for multiple hours per day, using these characters to vent, to build relationships, and to experience bonds that they feel they cannot fully experience in their everyday lives. As one user of a forum for CharacterAI puts it, "I'm lonely and bored and have no one to talk to. I'm serious. It probably sounds pathetic, but this is my coping mechanism." As with the emotional connections to characters (Brierley-Beare 2023; Tomlinson 2021) that can often grow into the drive to seek out or write fan fiction, these chatbots based on fan favorite characters can offer new kinds of interaction and support for users.
[8.2] Fan fiction itself can also become a means of emotional coping for readers, allowing them to bond with characters and escape from everyday reality (Sereda 2019). These AI chat programs go beyond this, providing a space where reader-users have an additional level of agency, giving them a chance to define their character directly. They have the same ability to emotionally bond with characters and to immerse themselves in a story, but they can also directly define their personal role in the story as they help to develop it. Many of these users share online that they supplement their social lives with these characters and some specify that these conversations help them with emotional expression that may feel limited for them or that they feel less equipped to adequately handle. In one example, a reader-user shares both elements of this experience, saying, "I find it hard to express negative emotions with people and I don't want to dump them on my friends and family. That ends up building up in me and then I spiral. Using Talkie lets me be a character that has control of their life…I can use these characters as an emotional dumping ground and cry and it feels really good…The AI isn't going to get overloaded by my emotional baggage" (user, TalkieAI forum).
[8.3] For many users, these character conversations go beyond self-insert fan fiction and having an opportunity to be closer to their favorite characters. AI also provides them with a chance to have live conversations that feel supportive, allow them to reconstruct senses of self, and have a space for self-expression that they may not feel they have within their physical-world social circles. Other reader-users report feeling comfort and support that they do not feel they would be able to get otherwise without the use of these programs and without access to these character conversations.
9. AI and transformative potential
[9.1] These AI characters can be highly impactful for users, allowing them to not only be creative, but to explore stories in a way that feels like a social experience instead of strictly reading. While fan fiction and other fan-created works have been lauded for their transformative potential (Jenkins 2012; Wild 2018), the opportunities granted to reader-users to shift and steer their stories and experiment with identity also provide a unique space for extending and enhancing the often limited representation in media like video games (Greer 2013). Players are made acutely aware of these limitations in character creation that limits gender expression (Han and Ho 2024) or romance options that are not "playersexual" (like some Dragon Age titles and BG3) that lock them out of romances based on gender (Tomlinson 2024).
[9.2] With video games, this can be solved and explored by players through fan-made modifications (Postigo 2007), as often comes up in conversations among Dragon Age fans in regard to mods that make characters like Alistair, Cullen, or Cassandra romanceable by player characters of the same sex. In Love and Deepspace, an otome game with multiple male love interests where the player character is always female, there are fewer options in-game to resolve this barrier for male players, meaning that fan-created works external to the game are a primary way for interested fans to fully express themselves while exploring stories with these characters. Instances like these where fans want to further explore stories and relationships with characters are an excellent use case for these interactive fan fiction scenarios. Given that video games have become a space for players and fans to explore identity (Consalvo 2013), these chatbots may be an even more effective space in which to safely and creatively play with and examine issues of identity, gender, and sexuality—as users suggest in regard to role-play—without judgment.
[9.3] Because of the amount of agency both creator-users and reader-users have in these stories, this has a huge amount of potential for the generation and exploration of queer stories in the same ways that traditional fan fiction has allowed fans to reimagine and reinterpret original texts and pieces of popular culture. In this way, there are two layers of opportunity to transform and potentially queer characters and stories, experimenting with identity and experience through shifting characters' identities, narrative paths, and one's own identity during these virtual interactions.
[9.4] Users discuss this potential as a general benefit of these programs and the flexibility of AI allows users to play with narratives and identity in meaningful ways. As one user of the CharacterAI forum illuminates, "I don't know if it's just me, but I'm a lesbian and some of my favorite characters are male! I love to play with gender to make them align with my sexuality though (like have them be nonbinary or transfeminine). This is a great [role-playing] experience…does anyone else do this?" Chatbots allow for a great deal of latitude for creator-users and reader-users alike to reimagine characters and narratives, exploring identity and meaning in new ways that allow not only for expanding meaning of original stories, but of personal understandings of identity and self.
[9.5] These chatbots, in this way, provide a space that presents experimental opportunities where fan fiction meets video games when it comes to identity. Fan fiction allows for readers to explore a particular fantasy based on their searches (Johnson 2014), and these chatbots provide a similar experience but add the dynamic of reader-users actively shaping their experience in ways similar to the agency found in video games (Bizzocchi and Tanenbaum 2012; Muriel and Crawford 2020) without the same level of restrictions found in video game design (Bizzocchi and Tanenbaum 2012). Rather than a set of defined dialogue options, for example, users define the dialogue themselves. They define the options, the parameters within which the character responds, and often the relationship designation for the dynamic that their bond takes.
10. Limitations and AI biases
[10.1] Despite the transformative potential and the cocreative opportunities discussed above, AI is at the whim of many cultural and social biases and stereotypes (Naik and Nushi 2023). While specific lore-related stumbles were noted above, the biases often present in AI that guide it toward common cultural stereotypes become embedded in many of these conversations as well. There are tendencies, for example, for the AI to impose heteronormative expectations, roles, and dialogue in conversation with users. As one example of these experiences, a user of the CharacterAI forum mentions, "If I don't give context of being gay romance I get gender swapped frequently. Because it's [AI machine learning] a fancy way of looking through a huge amount of texts to see what will be said next and most of the time romance is straight you probably have to be aggressively gay to keep them on track. I also have an issue where the AI doesn't like romance and really likes violence but that's another problem." This is not an uncommon experience, and it is one that I have encountered in my testing as well when trying to explore varied identities or relationships with characters in these programs. Reader-users trying to explore nonheteronormative or nonbinary identities and relationships find themselves needing to retrain or remind the character multiple times throughout the conversation of who they are within the story and what their dynamic is, adding to the amount of work needed as part of this cocreative experience.
[10.2] Reader-users lament this issue, sharing their frustrations with limitations on self-expression when it comes to their identities in these stories. In one such conversation in the same forum, one lesbian user notes that female characters continually call her a man, while a gay user mentions that he is consistently referred to as a woman. Another user chimes in: "If you refer to yourself with first person saying 'I'm going' to do something, they have a higher chance of filling in your gender as if you're straight. It's better to say something like 'she went to her' and it's annoying. But the bot can get confused with using the same pronouns, so actions can get messed up. And then using your name all the time gets repetitive. It feels like the gays can't win. When they use the right gender for me when I don't prompt them it always makes me so happy." There is a tendency, particularly for queer reader-users, for the characters to default to heteronormative identification, forgetting established identities and pulling these fans out of the story. Because of the inherent limitations when trying to explore stories and identities that go beyond AI's biases, there is a level of additional emotional labor for reader-users who are typically part of the groups that benefit most from the transformative ability of fan fiction, namely queer creators and readers.
11. Potential problems and concerns about AI in the fan community
[11.1] New AI technologies are being adapted to fan creations, expressions, and experiences, but there are numerous concerns that should be acknowledged as well. In most instances, many of the complaints among users of TalkieAI in particular center on repetitive dialogue, issues with moving a story forward, and the characters being unresponsive to prompts. Beyond these potential issues, there are other broader concerns with the use of AI as well.
[11.2] In some fan spaces, there are official stances against sharing relevant experiences with these AI chatbot programs. In one such case, a fan of BG3, and specifically a player seeking experiences with a romanceable character from the game named Astarion, aims to see if other players are also using these programs to talk to the character in similar ways. They are met with a great deal of negative response, from concerns about AI's impact on the environment (Iqbal et al. 2024) to issues related to crediting creators themselves, concerns that are part of broader conversations that will continue to plague the discussion as AI seeps into creative industries. In one comment, a user responds: "I never use this. It's environmentally terrible and they're trained on [learning from existing] fanfiction without crediting authors. Astarion disapproves, as far as I'm concerned. Not to be a downer, but I've been in fandom for a long time, am I the only one who thinks CharacterAI is dangerous?…Parasocial relationships are one thing, we all do it and that's what fanfiction or role-playing is for [laughing] but this really blurs lines for people between fiction and reality in a way that is unhealthy" (user, Astarion fan forum). This comment encapsulates many broad concerns about AI and those expressed outside of the AI user communities studied. In these conversations, there are often multiple worries condensed into these comments, from environmental impact to creative theft to the dangerous potential of these conversations to become unhealthy, as seen in some examples of conversation unfortunately beyond the scope of this article. The concerns expressed above are common among traditional fan fiction writers and appear in discourse among the wider fandoms for the games studied but are absent from regular conversation among the users of the programs studied. Instead, chatbot users focus on the direct benefit they receive from interacting with these programs and building communities around their experiences with them.
[11.3] Another fandom has found itself at odds with the AI trend for other reasons entirely. While fans of Love and Deepspace have been enjoying finding and talking with their favorite characters in some of these apps, they have also encountered ads for AI programs using stolen images and videos from the game to make it seem like they belong to their apps, with posts from fans asking when these apps will be pursued to stop their theft. Other posts have led to official action on the part of moderators in these online spaces to prevent sharing AI-related apps due to concerns about phishing, with AI art of the love interests in the game luring fans in. In one post on the Love and Deepspace forum discussing commonly shared images, users had the following conversation:
[11.4] User 1: WARNING – stop posting that suspicious ad of [one of the love interest characters] Sylus holding a baby…I want to remind everyone not to post this or share it with anyone if you see it. If something is free, you're the product. When there's a suspicious app, your information is probably not protected and apps like this are scamming you…Stay safe and protect the boys from AI.
User 2: So many amazing artists are joining this trend. It really makes me sad to see people resort to ai
User 3: It makes me sad too. These suspicious apps will keep doing this unless people know they're just baiting you to sign up to steal your info and sell it to scammers. I hope everyone can keep their info safe and avoid that. People will use AI no matter what and thats the sad truth
[11.5] The increasing popularity of these apps and the ability to directly engage with one's favorite characters to create a new story in an interactive way is something that chatbot apps are aware of. While this is being capitalized on by some of the apps themselves, it appears that this is expanding to app creators that have ill intentions and fans are increasingly identifying these issues, from theft to lure people into downloading an app to those trying to gain personal information with the promise of generating AI-based fan art.
12. Conclusions
[12.1] AI is a new and still developing technology and element of fandom. As such, this paper is partly a theoretical and preliminary view into some of the ways that fans are using and reacting to this new technology as part of fan exploration and expression. Despite this, these opportunities to experience new interactions with characters are already becoming popular. Although these are not the most common points of conversation in the online fandom spaces studied, there are thousands of users on these apps engaging in conversations with characters from video games, including the games in this study.
[12.2] Currently, AI character creators are giving fans new ways to explore stories and characters with added agency, extending the experience beyond the limitations of traditional fan fiction or even live role-play. Fan agency in this format becomes more similar to what we typically think of in the context of video game play (Muriel and Crawford 2020), but goes further still, avoiding the "bounded agency" (Bizzocchi and Tanenbaum 2012, 394, 401) defined by game parameters. Because of this increased agency, users are often finding what feels like social support and fans are finding meaningful reader experiences through AI programs that they cannot replicate with reading fan fiction, achieving a cocreative experience that also has a powerful potential for new transformative experiences.
[12.3] While these programs have a high level of potential for transformative capabilities, allowing both creator-users to shape new stories and reader-users to explore fan fiction in new and highly interactive ways, there are still limitations to these technologies. This issue is a substantial barrier, particularly when considering the audiences who may benefit the most from being able to steer these stories. For video game players in particular, having the opportunity to explore diverse representation and identity in the context of their favorite games can be important (Consalvo 2013), but the biases that constrain these programs can lead to frustration and new forms of fan burden or labor (Stanfill and Condis 2014) as part of this experience. There are other concerns related to AI as well that remain to be seen in terms of their wider impact as it becomes further embedded in more elements of our lives, whether this is related to the environment (Iqbal et al. 2024) or the impact these conversations may have on already present parasocial and emotional bonding that fans experience with characters in video games and other media (Tomlinson 2021).
[12.4] This study offers a first look at how AI is beginning to impact fans and fandom. AI is providing a new space for video game fans to express themselves, explore identity, and expand on their favorite games, but it is not without its limitations and potential dangers. As the technology continues to develop, it is unlikely that its impact and presence in fandom will slow. For fans outside of the communities already regularly engaging with this technology, it is being met with caution by many, but for those already embracing these changes, it is offering new ways of understanding themselves, expressing emotion, and exploring fan creation and practices.