‘NO TO AI GENERATED IMAGES’: Fan art creators contesting AI integration on social media platforms

Economic perspectives: The profit problem

To begin with, the precarious economic status of fan art creators can be described as, ‘operating from a position of cultural marginality and social weakness… and lack[ing] direct access to the means of commercial cultural production’ (Jenkins, 1992: 26). Fandom has long been considered a productive practice, and the media industries have started to recognize the economic value of fan creation and to ‘take fandom seriously’ (Fiske, 2002:41). While fan-produced content generates significant value for media producers and marketers (Stanfill and Condis, 2014), fandom communities continue to experience systematic ‘appropriation and exploitation for value as data’ (Couldry and Mejias, 2019: 338). Taking the Chinese entertainment industry for instance, media corporations take popular online fanfiction over for commercial adaptation and produce web series, which has generally become a highly popular and profitable media genre in the mainstream entertainment market (Wang and Wang, 2023). However, fandom – the infrastructure for content recreation and circulation – still largely remains uncompensated (Stanfill and Condis, 2014; Wang and Wang, 2023).

The global creator economy is fundamentally shaped by the asymmetrical power relations between individual creators and platform companies (Cunningham and Craig, 2019). Creators are generally aware of the potential commercial viability of their online content that can be exploited by the platforms. Like many creators, fan art creators also seek to establish clear boundaries between non-commercial content sharing and commercial exploitation of copyrighted material (Meese and Hagedorn, 2019). Fan art creators often regards reusing their work as fair use, ‘at least not harmful and may help sales of authorized works by increasing loyalty to the source’ (Tushnet, 2007: 144). However, as Tushnet (2017) observed, non-commercial fandom cultural production now faces unprecedented challenges in preventing their free-sharing content from being appropriated by platforms for commercial use. In particular, Chinese social media platforms efficiently datafy fandom content and convert fan engagement and creative labour into monetizable data traffic (Yin and Xie, 2021). However, despite plenty of scholarly discussion on how social media platforms capitalize on fandom communities, fan art creators – who are not only fans but also platform-based creators that strive to gain visibility and rewards for their work – have received limited attention (Pokrovskaya and Gronic, 2024). With the integration of AI systems by social media platforms that can extract creators’ work for lucrative use, further discussion is needed to examine the impact of these changes on fan art creators.

In addition, although fan art creator mostly creates content for non-commercial purposes, they sometimes try to monetize their content as a source of income (Pokrovskaya and Gronic, 2024). Besides in-app tips from fans, fan art creators often monetize their work by posting contact details on the platform to get business opportunities outside the platforms, such as providing customized illustration commission and selling self-made merchandise, such as posters, zines and accessories that featuring their derivative work (Pokrovskaya and Gronic, 2024; Sun, 2020). Therefore, gaining recognition and visibility for their works on the platform becomes the key of fan art creators’ monetization potential (Huang et al., 2023).

Though growing in popularity, fan art still very much exists in a sub-culture and legal grey area (Fauchart and von Hippel, 2008). Given fan art products made of unauthorized use of original work, fan art creators often hide themselves from the mainstream marketplace to avoid the notice of copyright holders and seek to monetize through personal connection or community-based activities, such as fan art market and small exhibitions (Pokrovskaya and Gronic, 2024). In the discussion of the creative economy, the small-scale informal art production function beyond formal, large-scale structures can offer valuable insights into how marginalized creators negotiate the tensions in the broader creative industries context (Lobato and Thomas, 2015). The challenges creators face in maintaining control over their content amid platforms’ AI integration are further exacerbated by the ambiguous legal status of their transformative works.

Legal perspectives: originality and the attribution problem

Fan creators occupy a precarious legal position that mirrors their social status: marginal and tolerated rather than celebrated as part of the universe of creators (Tushnet, 2017). While there have been efforts to strengthen the copyright protection of fan art in China, fan art creators still situated in a powerless position within the current legal system (Pokrovskaya and Gronic, 2024). Subject to pressure from corporations and legal restrictions, individual fan art creators often operate underground and remain out of sight (Freund, 2016). With the remixing nature of their content creation, fan art creators posit as a different type of author to the long-standing notion of ‘romantic author’, who creates original works as a result of their individual genius (Pappalardo and Aufderheide, 2020). In digital environments, issues surrounding the remixing fan art creation – particularly concerning ownership, authorship and copyright – become increasingly prominent (Freund, 2016).

The authorship status of fan-created works exemplifies what legal scholars describe as the Internet's capacity for ‘assaulting the distinction between mine and thine’ (Woodmansee and Jaszi, 1994: 26). While online platforms make content publication and circulation easy (Meese and Hagedorn, 2019), the challenge of achieving content visibility makes attribution central to creators’ rights. While social media platforms thrive on user-generated content, scholars have discussed platforms’ inadequate attribution systems and their role in perpetuating copyright infringement (Kaye et al., 2021; Meese and Hagedorn, 2019; Tan, 2018). Monroy-Hernandez et al. ‘s (2011) study of Scratch, a social media platform for remixing creators, illustrates how creators uphold the values of originality and develop manual credit-giving practices to deal with the platform's insufficient automated attribution systems. This also stresses the need for separate analyses of how creators encounter the attribution challenges brought by a specific platform, with distinctive characteristics and features (Tan, 2018).

Fandom operates based on implicit social norms that are held in common by community members, what Fauchart and von Hippel (2008) termed ‘norms-based IP system’. Within the close-knit communities, members establish shared norms regarding copyright, such as encouraged creative practices, anti-copying rules and proper attribution practices (Darling and Perzanowski, 2017). As such, fandom community members established informal rules that they impose on each other to self-regulate issues, such as requirements for non-commercial use and proper attribution (Fauchart and von Hippel, 2008; Fiesler, 2007). Among fan art creators, an expectation of proper attribution – crediting the contributions of others – is nearly universal. Plagiarism – the unattributed copying with the expectation of receiving credit – remains one of the most serious offenses in fan art communities that typically results in public censure (Tushnet, 2017). However, AI model training requires scraping data from various sources on the internet, often including copyrighted works (Sokler et al., 2024). As platforms integrate AI systems that may act as automated plagiarists, it becomes essential to further explore critical questions about the rights creators should hold over their content.

What also makes proper attribution central to fan art creators’ rights is the challenges they face in asserting the originality of their work. Fan art creations are transformative and can demonstrate remarkable originality in revealing new possibilities within the existing intellectual property (Fiesler, 2007), typical examples include illustrations or fiction that depict the pre-existing characters with different perspectives, interpretations or artistic representation. Nonetheless, their work remains obscure and unpublishable as ‘fanfiction is often so deeply embedded within a specific community that it is practically incomprehensible to those who don’t share exactly the same set of references’ (Tosenberger, 2014:5). Fan art creators seek recognition from their peers and distinguish themselves from pirates by expanding new aesthetics and twists to the source materials (Tushnet, 2007). The subtle and subjective artistic styles of creators face significant uncertainty when confronted with generative AI, which allows anyone to produce replicas that mimic an artist's style and can be executed by interchangeable workers (Chia, 2022). It is therefore important to investigate the cultural implications of anti-AI movement, particularly regarding the value and artistic identity that creators uphold.

Cultural perspectives: Community norms and creator identity

Community is at the heart of fandom culture. Fandom communities cohere around shared appreciation for certain cultural products (Wang and Wang, 2023), with fan art creators producing work to share with community members based on their collective understanding of the cultural products (Sarikakis et al., 2017). As Fiesler (2007) pointed out, instead of seeing them in isolation, fan art creators should be studied as part of pre-existing communities with their own traditions of participatory culture. Fandom communities often establish norms and foster an exclusive notion of ‘member status’ (Sarikakis et al., 2017: 12). Alongside the uniformity of consensus, the controversies and tensions that are contested and iterated through are also their collective experience of shaping the shared norms (Tang, 2023).

The proliferation of new technologies has given rise to a new form of participation in which machines play a significant role as actors within communities – what Li and Pang (2024) termed ‘human–community–machine interactions’. In fandom members’ networked socio-technical interactions with AI, the ongoing negotiation has an impact on the community as a whole, such as shaping or reinforcing certain norms (Li and Pang, 2024). Recently, there has been a growing body of literature on AI and fans’ creativity in fandom studies. For instance, AI technologies like chatbots have enabled humans and non-humans to co-create, which introduces new forms of participatory fandom (Lamerichs, 2018). However, while AI may democratize fandom content creation, it also raises concerns about how its introduction could deepen existing inequalities, posing ethical questions about the value of fan labour, and the exploitation of fan-made content (Lamerichs, 2018; Mussies, 2023).

The adoption of AI technologies in creative work has sparked debates about the human creators’ identity. Heated debates have emerged regarding the creative agency, subjective affection and experience of AI compared to human creators in cultural production (Chia, 2022; Messingschlager and Appel, 2023), echoing historical tensions between artisanal identity and industrial production during the Industrial Revolution (Hesmondhalgh and Baker, 2011). Recently, Ashton and Patel's (2024) study on the artistic identity construction of Ai-Da, the world's first AI robot artist, illustrates how its development team seeks to construct its creative worker identity through strategies such as showcasing artistic processes and cultivating audience relationships. However, further research is needed on how human creators receive and respond to these artificial creative identities. As Tang (2023) suggests, ambiguities of fan identity negotiation are intertwined with how norms are being violated, making it crucial to investigate how fan artists (re)negotiate their creative identities in response to the platform's AI integration.

Research methods and data collection

To collect protesting posts, we conducted targeted keyword searches on Lofter and identified the hashtag #ANTI-AI as the dominant hashtag creators used in the protesting movement. Lofter users also used both hashtags #Against AI (反AI) and # Protest AI (抵制AI) aside from #ANTI-AI in their protesting posts. However, as the prefix – ‘ANTI’ – in Chinese refers to either ‘go against’ (反对) or ‘protest’ (抵制), we thus focused on the hashtag #ANTI-AI with the specific Chinese word ‘抵制’ (protest) to minimize posts linking multiple hashtags to boost exposure.

The anti-AI protest went viral after Lofter's announcement on its launch of the AI drawing tool on the 6th of March 2023 (Li, 2023). We, therefore, collected all posts under the #ANTI-AI hashtag from March 2023 till October 2023 to capture the timely responses from Lofter users. The posts were manually exported to Excel for coding due to the platform restrictions. To protect the privacy of Lofter users, identifiable personal information, such as username and contact details, was deleted. Posts that were not directly related to Lofter's launch of the AI drawing tool, such as posts that only add the #ANTI-AI hashtag for boosting content exposure, were excluded from dataset. In total, 648 posts were included in the analysis.

We started to code the posts from the most liked and commented posts independently. The authors reviewed coding results at every 10 posts and held regular moderation meetings to maintain consistency and minimize potential bias. At the number of 50, there was no new theme emerged. We continued to code another 50 posts and reviewed the remaining posts together to ensure that data saturation was reached. The coding process was iterative, employing both deductive concept-driven and indicative data-driven qualitative content analysis approaches (Schreier, 2012). The two authors first developed three categories for coding in the codebook – Fandom Community and Culture (Jenkins 1992; Li and Pang, 2024), Author Economic Precarity (Fiske 2002; Tushnet 2017) and Attribution and Authorship (Kaye et al., 2021; Meese and Hagedorn, 2019), based on existing research and literature. During the coding process, we refined the codebook and expanded the categories into six to account for a fuller depiction of creators’ concerns. The refined categories and a summary of the coding statistics are shown in Table 1. The units for coding within posts varied. For example, some posts contained content that linked to multiple categories, while some other posts only contained an image with the hashtag. Each post was coded based on its content, meaning a single post could receive multiple codes. As a result, the total count of codes presented in Table 1 exceeds the total number of analyzed posts (n = 100).

OPEN IN VIEWER

Table 1.

Statistics and key concerns raised by Lofter users.

Categories	Numbers of frequency	Key concerns
Protection of community norms and culture	20	The loss of Lofter as a hub for the fan art community and its unique norms.
Protection of originality and attribute responsibility	47	Lack of originality in AI-generated content; AI-generated content as plagiarism.
Threats to economic gain	44	Limitations on creators’ autonomy in pursuing income opportunities within or outside the platform.
Creators returning to Lofter	57 fan art creators returned and 15 users stayed on the platform without leaving (out of 90 accounts)	Challenges of platform migration due to emotional attachment and established networks on Lofter.
Quoting from other Anti-AI movements	9	The call for all creators to unite in resisting AI-generated content.
The conflict between Anti-AI groups and AI supporters	3	Hypocrisy: Criticism of users who join the anti-AI movement while featuring AI-generated content on their personal pages.

Findings and discussion: When fan art meets generative AI

As we will discuss in this section, we identify three key themes in the anti-AI protest, revealing creators’ concerns about how the integration of AI on Lofter could pose a threat to their economic gain, the reputation of their original art creation and sub-culture community norms.

Implications and conclusion

In conclusion, this study investigates the multifaceted challenges faced by fan art creators through examining the grassroots anti-AI protesting movement; and more importantly, it foregrounds the underlying tensions that arise from platform-based AI integration in relation to this creative community. Our findings suggest that fan art creators represent a salient case for examining the asymmetrical power relations between individual creators and platforms in AI integration. Fan art creators feel at a bigger stake in the ongoing debate about AI's takeover of human artists, as platform's introduction of AI technologies has placed them in a dual tension: On one hand, creators have historically struggled to assert their rights over their work, as control over cultural products has largely remained in the hands of major media corporations (Sarikakis et al., 2015); On the other hand, platforms’ top-down integration of AI leaves creators with little choice over their online content, exposing it to the risk of being appropriated as AI training data. Fan art creators regard generative AI as an extension of platform capitalism, which reaps unparalleled benefits while diminishing their creative autonomy. In addition, fan art creators have expressed concerns about AI technologies appropriating their unique artistic expressions without proper attribution, which poses significant challenges to creators’ recognition, public reputation and the integrity of creative communities. Moreover, the unwelcomed trespass of AI tools on creators’ communities undermines the shared values and norms that fan art creators cherish, which are fundamental to their artistic identity. The creators’ dependence on platform-specific networks for visibility and audience engagement further complicates their autonomy to migrate to alternative platforms. This precarious situation underscores the vulnerability of sub-cultural creators who are embedded in niche communities that lack broader support outside of these platforms. We suggest that our empirical findings from fan art creators therefore have implications for understanding the perils posted by generative AI to individual and grassroots creators, and what rights do, and should creators have over the content they generate.

Besides the industry hype around the positive potential of generative AI in creative work, this study reveals the need for a more consensual environment in the integration of AI. The platform's apology and removal of AI features did not address creators’ dissatisfaction, which indicates that platforms, as both intermediaries for and proprietors of AI tools, should build transparency and consent into systems by design. At a time when comprehensive guidance from copyright law or platforms concerning AI remains lacking, this study demonstrates that creators are beginning to develop norms and practices around the use of their online content with regards to AI. In addition, we suggest that the nuanced norms and controversies emerging within specific groups around AI can inform scholarly research and debates on copyright systems (Fauchart and von Hippel, 2008), which should change to better account for everyday media practices of ordinary creators alongside the large-scale industrialized creative production (Meese and Hagedorn, 2019).

In the coming years, concerns around creators’ rights will be exacerbated by social media platforms’ quick embrace of generative AI tools. Shortly after Lofter's anti-AI protest, China introduced the first administrative policy on regulating platforms’ incorporation of AI, specifying the requirement for platforms to attribute and watermark AI-generated work (Cyberspace Administration of China, 2023). In December 2024, the UK government initiated a consultation aimed at clarifying how creators’ copyrighted materials can be utilized in AI model training, seeking to ensure that creators maintain control over their works while fostering AI innovation (UK Government, 2024). While the creators in our dataset agreed that AI poses multiple challenges to their rights, what they want from the platform or how they negotiate the tensions within the industry remains an open question. How to build a consensual environment, as well as the trust between the creative and tech sectors, merits further investigation.