Abstract
The adaptive evolution of digital platforms and their design changes remains understudied and platform change often remains obscure. The aim of our study is to unravel changing platform design in a process that we call adaptive governance: the process of strategic, platform-directed design change in response to internal and external stakeholders. We identify three strategies. Via capture, a digital platform introduces features that take over functions produced by users. Via imitation, platforms mimic the functionalities of other digital platforms. Via alignment, they implement features because users demand it. Our study therefore relates to the platform governance literature. Strategies such as capture and imitation represent governance by platforms since it shows how Twitch has an influence on its users. On the other hand, the strategy of alignment is about governance of platforms by showing that external actors can also have an influence on their design. We test our framework in the case of Twitch. We tracked design changes on the digital platform with a mixed-methods approach, combining analysis of chroniclers, interviews, and content analysis. Hence, we reconstruct a picture of adaptive governance on digital platforms by showing that Twitch has developed through a continual adaptive process of capturing user practices, imitating competitors, and aligning to meet user demand. This framework can be utilized in future critical studies, enabling researchers to identify pressures and patterns motivating platform design change.
Introduction
From July until October 2021, streaming platform Twitch had to deal with a surge of hate raids; the rapid influx of accounts posting hateful messages. As a response, streamers organized an online campaign (Han et al., 2023) termed #TwitchDoBetter. They organized the protest #ADayOffTwitch, in which streamers would not stream on the platform for a day. As a response, Twitch increased its moderation tools so streamers would be safer on the platform. Hence, the pressure by users led to design change. Unfortunately, this type of adaptive evolution of digital platforms and their design changes remains understudied and platform change often remains obscure (for exceptions, see Helmond et al., 2019; Partin, 2020; Van Der Vlist et al., 2022). However, a platform should not only be seen as an ensemble of technical elements, but also as the relational intersection of multiple stakeholders (Helmond et al., 2019; Scharlach and Hallinan, 2023). The aim of our study is to unravel changing platform design in a process that we call adaptive governance: the process of strategic, platform-directed design change in response to internal and external stakeholders.
We argue that three mechanisms are relevant to adaptive governance: capture, imitation, and alignment. Existing research has mostly focused on one of these processes; capture (Partin, 2020). Via capture, a digital platform introduces features that imitate functions produced by users. However, as our example shows, we argue that there are two other processes of platform change that remain a blindspot in the literature so far. First, via imitation, platforms mimic the functionalities of other digital platforms. A recent analysis of Twitter considers imitation between platforms as part of a larger “intermediality” driving platform change (Scolari, 2025); we argue imitation specifically is central to platform evolution. Second, via alignment, platforms implement design changes because users demand it, as can be seen in the example of #ADayOffTwitch. Capture, imitation, and alignment represent platform responses to environmental pressures. Existing platform studies typically contextualize platform change in relation to specific events (Gerlitz and Helmond, 2013; Hemphill, 2019; Highfield and Miltner, 2023; Mahetaji and Nieborg, 2023; Tarvin and Stanfill, 2022) but overlook the broader institutional context that digital platforms are embedded in and how this environment shapes their strategic decisions (Marchal et al., 2025). We advocate for a macro-level description of strategic platform change that identifies pressures (and consequent decisions) experienced by all platforms. Our aim is to contribute a testable framework for categorizing platform change.
For our empirical study, we test our framework on the case of Twitch, which is a website hosting thousands of live streaming channels created by users. The majority of channels are managed by hosts who broadcast visuals from a body/face cam and microphone alongside video feed (for instance, of a video game that they are playing). Other users can tune in and send messages to the host in real time, who reply and carry on multiple conversations at once. Over time, Twitch has implemented a plethora of features. By tracking these changes in design with a mixed-methods approach (combining analysis of so-called chroniclers, interviews, and content analysis), we reconstruct a picture of adaptive governance on digital platforms.
Our contribution is also methodological. We introduce the methodological concept of “chroniclers”: content creators who focus on documenting announced and observed changes on platforms, and interpret the potential impact and meaning of these changes for viewers. Chroniclers are more developed versions of vocal, non-expert online users able to influence public discourse, previously dubbed a “commentariat” (Highfield, 2013; Joseph and Bishop, 2024) or a “viewertariat” (Ampofo et al., 2011). Twitch has continually added, removed, and changed large portions of its functions since its start in 2011. So significant are some of these changes that content creators have chronicled them, predicted potential outcomes, and speculated on Twitch's intentions in making these changes. By analyzing how content creators talk about digital platforms over time, we can assess how these changes were perceived and infer what the platform's motivations were. Hence, we present a novel way of analyzing platform design change. Digital platforms have a “moving target” problem since platform change often remains obscure. We show that we can circumvent this problem by analyzing chroniclers who provide us with snapshots of design changes. Given chroniclers’ tendency to use and study multiple competing platforms, this method can be useful for analysis of other platforms as well.
Theoretical framework: Adaptive governance of digital platforms
We define platform governance as “the set of legal, political, and economic relationships structuring interactions between users, technology companies, governments, and other key stakeholders in the platform ecosystem” (Gorwa, 2019a: 2). Within platform governance, one can make a distinction between governance by platforms and governance of platforms. Governance by platforms refers to the political effects exerted by platforms (Gorwa, 2019b). For instance, digital platforms take on the responsibility of moderating content and policing the activity of their users (Gillespie, 2017). Governance of platforms refers to how other actors have an influence on them (Gorwa, 2019b). For instance, policies can emerge that specify platforms’ liabilities for user content and hosted activities. Both the ways that platforms govern their own users and the ways in which regulatory frameworks are imposed on platforms affect how online speech is governed (Gillespie, 2017). For example, Ahn et al. (2023) show how Facebook did not only change its content moderation policies by itself, but also as a reaction to political and legal contexts in the United States, Germany, and South Korea.
As we will show, adaptive governance interacts with both these dynamics. On the one hand, strategies such as capture and imitation represent governance by platforms since it shows how a digital platform has an influence on its users. On the other hand, the strategy of alignment is about governance of platforms by showing that external actors can also have an influence on design. This means that we do not conceptualize governance of platforms as mere regulation by government actors (as is the dominant approach in the literature); any external actor that influences platform design engages in its governance. Hence, in adaptive governance, the dynamics of governance by and of platforms interact.
Existing research on how platforms change their design has started to recognize that “outside stakeholders seemingly exert considerable pressure on platform companies which can spur change” (Barrett and Kreiss, 2019: 14). Multiple terms related to governance have emerged without a clear framework to explain the relationship (if any) between them. Barrett and Kreiss (2019) show how pressure from journalists influenced changes in the “I’m a Voter” button on Facebook, dubbing rapid platform changes as “platform transience”. Ananny and Gillespie show how platforms enact disruptive changes (“shocks”) and, following public outcry, reverse the changes in what they call “soft governance.” User protests over changes is “insufficient as a form of platform governance, by themselves, but we believe they could be extended” (Ananny and Gillespie, 2016: 3). YouTube's public response to criticism over its inconsistent content moderation has been dubbed “governance-washing” (Tarvin and Stanfill, 2022). Marchal et al. (2025) show that negative media coverage of Facebook, Twitter, and YouTube—as well as competition between platforms—has a direct effect on changes in their user policies. Burgess and Baym (2020) demonstrate how practices by users (such as the @ feature) were incorporated (captured) by Twitter. Turvy and Abidin (2025) identify a “patchwork governance” of community norms and official policies governing the TikTok's evolving approach toward child influencers. Additional work has suggested understanding platforms through infrastructural (Ananny and Gillespie, 2016; Mahetaji and Nieborg, 2023), political-economic (Poell et al., 2021), or organizational studies frameworks (Scolari, 2025). So far, the field of platform studies has introduced case-specific terminology, or broad frameworks informed by other fields. What is needed is an approach that links specific platform changes to a holistic framework.
In order to fill this gap, we argue that digital platforms change their design based on a process we call adaptive governance: the process of strategic, platform-directed design change in response to internal and external stakeholders. Internal stakeholders are those actors that are already embedded in the platform, for example, content creators and users. They regularly and intimately interact with the platform and its content. Their roles as creators and users are defined by their presence on the platform. External stakeholders are more distant to the platform, such as governments, other digital platforms, advertisers, and journalists. Although they do not intimately and frequently interact with platform activities, they still are impacted by the platform or have a demand for design change. The lines between an internal and external stakeholder are not always clearcut, but it is meaningful to distinguish between those actors that are more embedded into a platform and those that are more distant since this distance allows for different incentives and pressures for change.
Hence, platforms acquire new features through a complex interaction with other actors in a platform ecosystem (Partin, 2020). This process affects both the technological infrastructure of platforms (e.g. algorithms, API), and the design of the interface between end-user and platform (e.g. buttons, visuals). Therefore, we adopt a broad understanding of platform design. Through the exchange, ordering, and management of data, features mediate and set the terms of connection between users, content, and digital platforms (Gorwa, 2019b; Lessig, 2006). They provide the key affordances of any platform (Burgess and Baym, 2020). Platforms change their design during adaptive governance, which comprises three strategies: capture, imitation, and alignment.
First, capture refers to a digital platform's practice of introducing features that imitate functions produced by users. It entails “entry by one platform provider into another's market by bundling its own platform's functionality with that of the target's so as to leverage shared user relationships and common components” (Eisenmann et al., 2011). By transforming user innovations into core architecture, formerly complementary relationships are transformed into competitive ones (Partin, 2020). When a platform introduces an in-house version of a feature, it becomes the “default” setting, making third-party tools feel redundant or inconvenient for users. It is about “platform owners who are parasitic on third parties” and “implies a kind of inevitability whereby the small are subsumed by the large” (Partin, 2020). This concept accounts for only some changes seen on platforms, however. We argue that adaptive governance includes not only capture of complementor ideas, but also mimicry of other platforms (i.e. imitation), and adjustment of features in response to user resistance (i.e. alignment).
With regard to imitation, platforms regularly compete with one another to attract limited users via attractive features (Scolari, 2025) operating as multi-sided markets that seek to become indispensable by leveraging their role as a network between users, partners, and developers (Helmond et al., 2019). We therefore adopt a platform ecosystem perspective that argues that platforms with superior technology are competitive on a market (Kazan et al., 2018). The need to innovate leads digital platforms to “tap” external sources, including competitors (Linder et al., 2003). Hence, platforms may be sensitive to “peer pressure,” as they try to align themselves with industry standards or to stay competitive with rival platforms that have made certain design changes (Marchal et al., 2025). In contrast to capture, which occurs within a single-platform ecosystem, imitation occurs between platforms, as competitors monitor and mimic features used by opponents. In the rapid rise of social media during the 2000s, start-ups (not yet understood as platforms) freely stole and borrowed features visibly popular elsewhere. According to Pardes (2020), this copycatting leads to a process of homogenization, by which digital platforms increasingly look like each other. Imitation between companies continues to be a great source of pressure on digital platforms’ decisions regarding design change and adjustment.
Lastly, alignment refers to the change of platform design because users demand it. So far, we have only assessed theories on how technology can shape users. However, users can also shape technology (MacKenzie and Wajcman, 1999). This social constructivist stance acknowledges that social norms have an influence on digital platforms and affect how users interact with technology (Mischaud, 2007). Hence, we should not lose sight of the multi-directionality of agency and connectivity between users and platforms (Bucher and Helmond, 2018). Alignment can take two forms. In the first version, users are significant for platform change, as by routinely interacting with a digital platform on a daily basis, they can motivate structures to change (Orlikowski, 2000; Siles, 2013). Via these cultures of use, platforms are given meaning by everyday life and practices of diverse individuals and communities (Burgess and Baym, 2020).
In the second version of alignment, users demand change. Existing research acknowledges that public pressure can work as a mechanism that propels design change (Marchal et al., 2025). An aggrieved community can engage in tactics traditionally involved in social movements: “seeking more visible venues for the expression of that outrage, expressing it in terms that implicate users beyond the aggrieved group, and delivering that outrage to the platform itself” (Ananny and Gillespie, 2016: 6). We argue that if users are dissatisfied with the workings of a digital platform and a critical mass is achieved, they can pressure the platform to bring about change in its design. This pressure can be in the form of, for instance, protest and media campaigns. On YouTube, this can take the form of “call-outs” by creators when they feel the platform has wrongfully enforced copyright (Hallinan et al., 2024). In some cases, platforms may simply appear to satisfy user demand without making substantial change, as in “governance-washing” (Tarvin and Stanfill, 2022), while in other cases features are genuinely shifted to appease users, as in “emergent governance” (Reynolds and Hallinan, 2025). When user-based pressure becomes too hard to ignore by the digital platform, it will implement the changes—that is, align its design—to address user demand. Even though digital platforms may change their design to show support for different causes and groups, the underlying logics of platforms—to maintain high user numbers and increase engagement and profit—remain (Helmond et al., 2019; Highfield and Miltner, 2023).
Methods
Case selection: Twitch
In order to demonstrate our theoretical framework, we selected Twitch as an illustrative case study of adaptive governance. Twitch operates as a platform, defined in Poell et al. (2021: 5) as “data infrastructures that facilitate, aggregate, monetize, and govern interactions between end-users and content and service providers”. Platform changes mediate networks of users (e.g. streamers and viewers) who, in turn, also mediate the platform when they adopt or resist the platform's changes. Synchronous, competitive, and built around active user communities, Twitch is an ideal case for examining the dynamics of adaptive governance.
Research design
Platforms change continuously, posing serious challenges to writing internet histories (Bayer et al., 2020; Helmond et al., 2019). In order to address this moving target problem on Twitch, we adopted a platform biography approach, which “generates a narrative of change by weaving together the stories of material objects, social relations, and events, and that therefore brings onto the stage the human lives that have intersected with and shaped the platform in question” (Burgess and Baym, 2020: 27). This biographical method begins with a feature and reconstructs the digital platform from that point, creating a narrative of change. It is not just the establishment of linear timelines, but instead a “complex, entangled approach that focuses on the relationships between media and other actors within the sociotechnological landscape” (Scolari, 2025: 656). Concurring that platform evolution can best be observed over a sufficiently long period of time (Helmond et al., 2019), we reconstructed the sources of change of Twitch's platform design from 2016 until 2023 using a variety of sources.
This study adopts a mixed-methods approach with three distinct strategies: (1) reviewing YouTube commentators as “chroniclers” of platform change on Twitch, (2) including qualitative interview and observation data, and (3) considering Twitch's official announcements via the company blog as an online archive. Twitch announcements were analyzed via a thematic analysis approach (Braun and Clarke, 2021), which identified underlying patterns in platform change processes. Via data triangulation of these sources, we could come to more reliable results. We analyzed data chronologically and focused on key design changes discussed across data sets (e.g. chronicler critiques, streamer interviews, and Twitch announcements). In this way, though data sources differed in intentionality (i.e. Twitch promoted changes while chroniclers criticized them), our data revolved around design changes that caused strong reactions among users. We interpreted these changes as indicative of a larger on-going governance strategy we characterize as “adaptive.”
We came to our framework abductively. Some work has been conducted on platform capture and envelopment, which we incorporated into our adaptive governance framework. Hence, we worked from theory to the empirics. At the same time, by analyzing and triangulating our different sources, we came to our other two strategies of imitation and alignment. Hence, the empirics and our knowledge of Twitch as a platform (e.g. one of the authors is a Twitch streamer herself) allowed us to theorize about our adaptive governance framework. Hence, in line with abductive analysis, we adopted “a creative inferential process aimed at producing new hypotheses and theories based on surprising research evidence” (Timmermans and Tavory, 2012: 170). The discovery of our theoretical framework depended on our inability to frame findings in existing theories and on our ability to extend existing theory in new ways (Timmermans and Tavory, 2012). Working from existing theory to the empirics, and from the empirics to theory, we came to our novel theoretical framework.
YouTube chroniclers
One valuable source of “archival change data” is the industry of YouTube content creators who focus on documenting announced and observed changes on Twitch, and interpret the potential impact and meaning of these changes for viewers. Previous work has centered YouTubers as critical subjects able to assess the affordances and limits of platforms (Bishop, 2020; Consalvo et al., 2023; Glatt, 2022). More recent work has understood YouTubers as practicing “user-generated accountability”, wherein “(…) YouTubers were critical of the platform (…) [and] sought to address these problems by making noise, or generating publicity to draw YouTube's attention to particular problems” (Reynolds and Hallinan, 2024). Similar dynamics take place between Twitch-focused YouTubers and the Twitch platform. Combining documentation with interpretation, Twitch “chroniclers” operating on YouTube (e.g. Zach Bussey, Senpai Gaming, Lowco) act as platform historians who narrativize platform changes as parts of larger stories in which creators, viewers, and the platform operate as co-dependent and expressive actors. For scholars attempting to understand platform decision-making without access to internal decision-making procedures, this folk history of Twitch offers years of notes on changes big and small, many of which are forgotten (such as boost trains) or assumed inherent later on (such as Charity tools or custom tags; discussed below).
Chroniclers produce an archive of commentary on Twitch's visible design changes; a semi-permanent, public record that is especially valuable when the platform erases past features. Chroniclers present a “window to the past” that allowed us to perceive platform design changes in that particular time frame. It enabled us to identify relevant design changes as well as understand how they were perceived at the time. Moreover, it allowed us to interpret the logic of the digital platform in bringing about these design changes.
We searched for chroniclers’ content in two ways. First, we searched for specific Twitch design aspects and moments on YouTube (such as hate raids, the introduction of Bits, the introduction of Twitch Charity). We selected those content creators that offered their own perspective on these developments, and watched and analyzed their videos. Second, for some larger creators who specialize in chronicling (e.g. Zach Bussey, Senpai Gaming), we went through their entire library of videos and watched them so that we could uncover perspectives that were relevant to our analysis. This also allowed us to discover new design changes that were not yet included in our analysis. We only focused on English-speaking chroniclers, which limits our findings. This is especially relevant on a platform such as Twitch where the user and streamerbase is transnational and multilingual. As can be seen in Table 1, chroniclers varied in size but shared a focus on live streaming, with the majority engaged with gaming and Twitch specifically. By including a wide selection of chroniclers, we could strengthen our claims about user perspectives.
Overview of the chroniclers included in the analysis.
Thematic analysis of interviews with streamers
Interview excerpts describing live streamer perception and understanding of Twitch design changes were included in this study, taken from a longitudinal study of streamers that included 34 interviews with 12 live streamers from March 2022 to August 2023. Interviewees were small variety streamers who create content primarily on Twitch, rely heavily on community contribution and are—by their own account—highly sensitive to design changes. Excerpts from these interviews present examples of streamers’ firsthand experiences with Twitch platform design, including experimental and discontinued features never fully integrated into the platform.
Streamers’ experience and knowledge of the platform is deep, and they have a high sensitivity to changes the platform makes; for instance, changes to the algorithm or shifts in content regulation and moderation. Sensitive to changes, outside of the regular platform company, yet operating in the front and center of the user–platform–creator nexus, streamers are expert observers of platform change. Related to this, one of the authors has been a Twitch streamer for 4 years, observing and experiencing many of the changes discussed in the Results section. Hence, the starting point of the framework was an insider's perspective (via interviews or streaming) on Twitch's design changes.
Analysis of official Twitch announcements
Two data sources are drawn directly from Twitch. The first is straightforward: documentation of current features listed in Twitch's guides for new streamers as well as the Dashboard feature. The second is Twitch's blog where major platform changes are regularly announced (e.g. the introduction of Cheering, Sound Bites, new moderation options). We focused on posts that were relevant to our analysis (i.e. that introduced new features or key feature updates on the platform). This meant omitting other types of announcements (e.g. TwitchCon dates, policy changes, promotional events). We analyzed the language used by Twitch to understand the intent behind introducing features. Taken together, this data provides descriptions of key design changes, an official timeline from Twitch, and occasional official explanations of the platform's intentions.
Our data sources—streamers, chroniclers, and Twitch itself—rarely agree on the usefulness of design changes. For many streamers, change is a disruption to their work, while for YouTubers it can be a source of content to mull over. For Twitch, design changes are business-as-usual. In combining these perspectives, we gained a fuller picture of which changes are most impactful based on how deeply they resonate with (or repel) key stakeholders. We do not offer an assessment on the “best” or “most successful” changes; rather, we identified patterns in the way these stakeholders interact via commentary on design changes that illuminates a consistent, recurring pattern amounting to an adaptive governance strategy by Twitch.
Results
Our analysis reveals that Twitch continually adjusts its design, which directly impacts content creators. Analyzing discourse surrounding these changes—including analysis of chronicler commentary, interviews with streamers, and close reading of Twitch's feature announcements—reveals a process of external pressures from users and competitors motivating Twitch to adjust its design in a process we call adaptive governance. We arrange our findings by adaptive governance type, beginning with Twitch's capture of user (i.e. complementor) added design, followed by imitation of other platforms’ design, and finally the implementation of user-demand into design in the process of alignment.
Capture
Capturing refers to Twitch's practice of introducing features that imitate functions produced by users, who become complementors adding to platform design. By monitoring their user base and third-party applications, Twitch gains knowledge about which features have a high demand. After introducing these features as Twitch features, Twitch can take a part of the revenue made from them. In general, “Twitch is taking a lot of ideas that have been around the community and making them first-party or making some direct monetization through them” (Gaming Careers, 2023b). We show this mechanism via three examples: Cheering with Bits, Sound Bites, and the Twitch Charity tool (for an overview, see Table 2).
Moments of design change related to the capture process.
First, in June 2016, Twitch introduced the Cheering feature, with which viewers can donate Bits (a virtual currency) to streamers (Twitch, 2016). Until then, other tipping services existed (e.g. StreamTips, Twitch Alerts, PayPal), for which streamers would get a larger cut of the donated money than when they donate via Twitch (Chadlantis, 2016). The donation market had been substantial and exercised through third-party services. Hence, Twitch wanted a larger share of the donation market (TotalBiscuit, 2016). One of the problems with services like PayPal was that viewers could recall their money (HybridPanda, 2016) and integrated services like Cheers are easier to use for viewers (TotalBiscuit, 2016). According to HybridPanda (2016), “PayPal has been used to send money through the internet for the longest time. So, Twitch is going to get something out of this, right? Right!”. Twitch does so by taking a share from the viewer when they buy Bits (about 30%, depending on the amount of Bits viewers purchase), plus transaction costs (King Gothalion, 2016).
Initially, the Cheering function was an option alongside third-party tipping services, but then Cheering replaced third-party options (Avoiding The Puddle, 2018). The fact that Twitch took a share of the donation faced significant backlash from streamers. For instance, according to Aris, “I don’t like bits. (…) Twitch is getting their fingers into the relationship between myself and fans of mine who want to donate to me. This doesn’t involve you” (Avoiding The Puddle, 2018). Regardless of this backlash, as of 2024, the Cheering function is an integrated part of Twitch culture. Twitch (2016) framed their capture strategy as an added value to the platform by saying that “(w)e believe that Cheering provides a completely new value to the Twitch community, and our goal is to take your feedback and make it a critical component to a broadcaster's success.” Twitch captured third-party methods of monetizing user interactions and then framed it as alignment with user interests.
Second, in 2023, Twitch announced the introduction of Sound Bites: first-party sound alerts that allow viewers to pay to interact with streamers (Twitch, 2023d). Twitch stated that this was in order to facilitate deeper community interaction and innovative viewer recognition (Twitch, 2023d). According to Gaming Careers (2023b), “(w)e’ve seen tools like this been [sic] developed for a long time (…), where they are (…) extension-based. Where somebody pays three dollars and they get a jumpscare or they get a certain sound alert (…).” Before capturing Sound Bites, Twitch even stated that sometimes it takes only one person to improve the Twitch experience (Twitch, 2018). According to Gaming Careers (2023b), the introduction of Sound Bites by Twitch is not surprising: Twitch observed that there was a significant amount of money in Sound Bites and a significant demand by users, and therefore decided to build a first-party version of the feature. As of 2024, the feature is a built-in Twitch extension called Sound Alerts.
Third, in 2022, Twitch (2022) announced the introduction of a charity tool, a “new built-in fundraising stream feature that allows you to set up and run a stream for charity in just a few clicks.”. According to Twitch, (2022) the tool is meant to reduce “(t)he hassles of running a charity stream”, such as tracking donations, converting subs and Bits, and managing multiple apps at once. In contrast to the Cheer and Bits system, Twitch does not take a share, nor enjoy tax benefits from the feature. Instead, it is a community feature that enables streamers to raise money for charity directly on Twitch without having to use other third-party services such as Tiltify or StreamLabs Charity (Zach Bussey, 2022). In interviews about her work, streamer Joanne was satisfied with third-party offerings: “Tiltify is what I've always used. I love it. … Sometimes I think like Twitch is trying to like reclaim stuff that's been made third party and like people are already like, ‘we don't need this’” (March 1, 2023, personal communication). Twitch is not necessarily competing with Tiltify in order to adapt its features, but rather identified a gap in Twitch features filled by a popular third-party extension. Subsequently, Twitch coopted a feature that was widely used by content creators. According to Lowco (2022), Twitch Charity “might get people's (…) toes dipped into the water of charity fundraising (…). Where maybe it was overwhelming for them to think about, now they can just do it with a few button presses.” Since August 2022, the charity tool is out of beta and available to all streamers (Twitch, 2022).
The example features of Bits, Sound Bites, and the charity tool demonstrate that Twitch monitors user behavior and user-made innovations, and then recreates in-house versions of their own. According to Partin (2020), “once third parties proved the viability of these services and began operating them at scale, Twitch was able to integrate similar tools into its platform, rebranding them (…) in the process.”. However, capture is not the only strategy of adaptive governance; we also identified the processes of imitation and alignment.
Imitation
The second adaptive governance strategy—imitation—contrasts sharply to this platform-user siphon. In imitation, Twitch mimics competing platforms. Imitation is not as final as capture, wherein larger platforms “capture market share by foreclosing an incumbent's access to users” (Eisenmann et al., 2011: 1270). When Twitch adopts competitors’ features, its users may be familiar with them, having used them elsewhere. Twitch monitors competing platforms as closely as it does its own innovative users, and readily emulates popular features (for an overview, see Table 3).
Moments of design change related to the imitation process.
Imitation is part of the so-called platform wars, wherein “behemoth” social media platforms can die quickly (Senpai Gaming, 2021). We provide four examples of this mechanism: Hype Chat, the Discovery Feed, Stories, and a 70/30 revenue split.
First, the Hype Chat functionality was announced in 2023. A copy of YouTube's existing “Super Chat,” “Hype Chat” allowed users to pin a colorful message to the top of a streamer's chat. The higher a donation, the longer the message would remain visible. Twitch announced later in the same month that it would discontinue the feature starting November 10, citing low usage (Twitch, 2023a). Commentary on the short-lived feature noted its unpopularity and feeling of redundancy among users as well as streamers’ displeasure at splitting revenue from Hype Chats 70/30 with the platform (Wolens, 2023). In its announcement, Twitch stated that existing Cheers and Bits functions would gain the ability to be “pinned” just as Hype Chats were. In this case, Twitch's imitation of a competitor was rapidly walked back and folded into existing features, effectively aligning with user demand.
Second, at TwitchCon Paris, July 8–9, 2023, Twitch (2023d), announced two features functioning in tandem: a “new discovery feature” that will “allow viewers to seamlessly scroll through a feed of recommended channels and clips with ease,” fed by streamers’ channel-specific “stories.” Discoverability on Twitch has long been considered weak, which streamer Nancy attributes to a lack of algorithmic guidance of users: “the reason why it's [discoverability] not good is because Twitch doesn't have an algorithm the way that other socials have algorithms that kind of suggest [sic] constantly suggest things to you” (July 26, 2022, personal communication). There was a demand by users and streamers for the Discovery Feed since TikTok and platforms such as Instagram used similar designs by which users could scroll through recommended content (Gaming Careers, 2023b). According to Gaming Careers (2023b), “(e)veryone is currently using Instagram and TikTok and YouTube shorts to post their Twitch clips. So if you can just do it natively on the platform … I think it will be huge for the platform.”.
The Discovery Feed allows streamers to take clips from their streams, and turn them into vertical content (i.e. content that can be edited and reposted to fit popular mobile apps such as TikTok). It allows for users to discover content in a feed even while streamers are not live (Hammer Dance, 2023). According to Lowco (2023), “(t)he reason that Twitch should and is getting into this, is because they want to make the platform bigger. And how do you make it bigger? Well, you tap into what other social media platforms are doing, that works really well, which is clips and short form content.”. According to Gaming Careers, it was “Twitch's version of TikTok” (2023a) and Senpai Gaming (2023) even called it “TwitchTok.”.
Feeding the algorithmically assembled Discovery Feed are “Stories,” which include clips, pictures, text, updates, and polls (Twitch, 2023b). This option for asynchronous posting directly addresses an issue streamer Mary Ann identified when interviewed in the spring of 2022: “When you end your stream, the world forgets about you, and that's really bad for a content creator.” (March 17, 2022, personal communication). According to Senpai Gaming (2023), Twitch Stories are exactly like features on other platforms, such as Instagram Stories, Snapchat Stories, or YouTube Stories. 1 According to Hammer Dance (2023), “this is interesting because Youtube (…) just recently started getting rid of (…) Stories. (…) But I feel like on Twitch this form of content might actually work really well, it's on a more personal. level.” In this case, users applauded Twitch's blatant imitation of competitors’ features.
Finally, in October 2023, Twitch restructured its revenue splits with streamers with the introduction of the Plus Program that allowed for a 70/30 revenue share (up from 50/50) offered to a minority of larger channels with high numbers of subscriptions (Twitch, 2023c). Smaller streamers are painfully aware of revenue split differences: “YouTube has always made more money than Twitch … most creators get like a 50/50 cut with Twitch. If you're really good, you can get a negotiated [sic] like 70/30 cut” (Mary Ann, August 8, 2022, personal communication). Twitch's re-introduction of a 70/30 split for certain streamers came shortly after competitor Kick offered higher revenue sharing. This is a strong indication of conscious imitative adaptive governance on the part of Twitch, as they try to hold on to major creators with greater revenue sharing.
While capture and imitation originate with Twitch observing and mimicking other actors (creative users and competitors), the final strategy of adaptive governance results from Twitch aligning its feature offerings to fit with user demand.
Alignment
Alignment refers to the implementation of features because users demand it. It is a quality of life improvement pushed for via a bottom-up process. Reliant on the platform for their income, streamers are eager to see improvements: “I just want—I want to believe that … it [Twitch] is making it more inclusive and that the trolls hopefully are discouraged. Like there's always going to be dark corners of the internet no matter what” (Joanne, October 2, 2023, personal communication). Twitch's slow response to popular user requests is frequently met with impatience: “(…) they [Twitch] are always so late. Again, it just takes [sic] to the point where everyone is (…) super unhappy and just like yelling and fuming for them to actually do something” (Lowco, 2021). We provide four examples of this mechanism: a list of three features that help with moderation and one feature that prevents copyright infringement (for an overview, see Table 4).
Moments of design change related to the alignment process.
First, as a response to large-scale harassment against especially LGBTQI+ people and people of color, a group of streamers organized the #ADayOffTwitch movement (also known as #TwitchDoBetter): a call for all streamers to not stream on Twitch the day of September 1, 2021. The main catalysts for this movement were hate raids; a redirection of viewers to a targeted stream in order to bombard their chats with hate speech and harassment. For instance, one of the organizers of the #ADayOffTwitch movement—RekitRaven—received messages such as “This chat belongs to the K K K” during her stream (see Figure 1) (ProgressiveLiberal, 2021).
A hate raid directed toward RekitRaven.
According to Lowco (2021), there is a lack of moderation and features that help streamers keep their communities safe, contributing to a toxic environment. RekitRaven said, “I’m going to call out Twitch (…) for not having proactive tools to prevent these things from happening. Obviously, hate raids are never going to go away (…). But you can also deincentivize it, make it harder to do” (Mike From PA, 2021).
During the #ADayOffTwitch campaign, the platform lost about 25% of their viewership. Viewership dropped from 4 million average concurrent viewers to just over 3 million average concurrent viewers (EsportsGG, 2021). Users’ collective demands were a roundtable discussion with affected creators, “proactive protection” (e.g. enabling creators to select the account age of chatters and deny incoming raids), a limit of three Twitch accounts to a single-email address, and an increase in overall platform decision-making transparency.
As a response to #ADayOffTwitch, Twitch responded “(w)e’ve seen a lot of conversation about botting, hate raids, and other forms of harassment targeting marginalized creators. You’re asking us to do better, and we know we need to do more to address these issues. That includes an open and ongoing dialogue about creator safety” (Twitch [@Twitch], 2021). In its alignment of features, Twitch introduced an update to better detect hate speech in chats (Parrish, 2021) and in January 2021, Twitch updated its Hateful Conduct and Harassment Policy to include the banning of doxxing (the publication of identifying information about a streamer) or hacking another person, as well as the malicious use of emotes (Twitch, 2020b). They also announced the launch of channel-level ban evasion detection and more options for account verification (Twitch [@Twitch], 2021). As of 2025, there are many more options available to streamers to prevent harassment and hate raids, such as a chat mode for followers only, sharing ban requests between streamers, and the option to manage incoming raids (Twitch, n.d.).
Commentators considered this alignment significant. Developing these additional tools implied that Twitch was willing to invest money, establish a more robust moderation staff, and take a stronger position on issues such as hate speech (Mike From PA, 2021). At the same time, the issue not only affected smaller streamers, but impacted larger streamers as well, which is likely why Twitch addressed it (NotSoHasanAbi, 2021). Twitch clearly considered these costly design changes worthwhile for maintaining a positive reputation among key streamers.
A second instance of alignment emerged over the fair use of copyrighted music during streaming. In 2020, The Recording Industry Association of America (RIAA) flagged songs via an automatic bot that went through all streamers’ clips and videos, and then issued automatic Digital Millenium Copyright Act strikes to Twitch. Twitch then issued strike warnings to these creators (after three strikes, streamers were permanently banned from the platform). This led to an outcry by creators, most visibly on Twitter (Devin Nash, 2020; Zach Bussey, 2020). Initially, Twitch advised offending streamers to remove the clips when they were unsure about the right holders to the audio (Twitch Support [@TwitchSupport], 2020), but this was difficult due to the large clip archives that streamers had. Addressing streamers, Emmett Shear—then Twitch CEO—confessed that “we should have had better tools ready to manage your content. (…) We wish we did. We’re sorry those tools weren’t available when you needed them and that so many creators had to delete their videos (…)” (Zach Bussey, 2020). Twitch's long-term solution was the introduction of a feature called Soundtrack by Twitch in September 2020, giving streamers a curated collection of rights-cleared music (Twitch, 2020a). Responses were mixed. Twitch quickly assembled Soundtrack by Twitch to prevent further problems with the RIAA. According to Senpai Gaming, “all they were actually doing was finding sneaky gray areas to circumvent paying proper copyrights, which made the RIAA extremely angry” (Senpai Gaming, 2020). Soundtrack by Twitch is a case of failed alignment; it was not widely used and closed down in 2023; there are now numerous third-party options to streamers.
Alignment of this kind appears to follow a certain pattern: “First, they [Twitch] acknowledge the issue, which is a good first step. (…) But then, second, (…) their solution is to put a bandaid on it. (…) It allows them to say, ‘hey look, we did something’, where in reality, it usually ends up making the problem worse” (Senpai Gaming, 2020). User demands tend to focus on quality of life issues, safety features, and increased customization, though it is clear pressure from non-user parties (e.g. the RIAA) that also triggers alignment by the platform.
Conclusion
Twitch has developed through a continual adaptive process of capturing user practices, imitating competitors, and aligning to meet user demand. These responses to external pressures amount to a form of governance we dub adaptive governance.
We contribute to the literature of platform change by identifying three practices that comprise adaptive governance. Via capture, Twitch imitates features that were produced by users. Via imitation, Twitch mimics competing platforms. And via alignment, user demand compels Twitch to change. These mechanisms can overlap; after user pressure for better safety tools in response to hate raids (alignment), Twitch implemented features previously innovated by users themselves (capture). Twitch's recent “Discovery Feed” imitates the “Stories” features of other platforms (imitation), and was announced as an answer to streamer demand for better discoverability on the platform (alignment). Hence, in identifying external pressures that cause platform-directed design change, adaptive governance brings together governance by platforms (where platforms exert influence on actors) and governance of platforms (where outside actors exert influence on platforms), contributing to the literature on platform governance (Gorwa, 2019a, 2019b). Importantly, we argue these processes operate as a push-and-pull dynamic between Twitch and external stakeholders, resulting in continual adaptation of the platform as a response to and anticipation of external pressures.
Our second contribution is methodological. A historical chronicling of social platforms can expand analysis beyond user experience and reaction to changes, and point toward underlying patterns in the platforms’ decision-making. Chroniclers can be understood as an advancement of the “online commentariat” concept (Highfield, 2013), valuable for their expertise but also their documentation of platform design changes that rapidly become forgotten. As vocal advocates, chroniclers have the capacity to define and promote the terms that user collectives pressure platforms to adopt, and their archive of work presents a more permanent record of platform design change. Twitch is not the only target of chroniclers; changes to Twitter/X, TikTok, and Instagram have been subject to lengthy critiques as well.
With regard to future research, we see three possible avenues. First, it would be interesting to see how the three different mechanisms of capture, imitation, and alignment can work in conjunction and reinforce each other. The example of hate raids is illustrative here. Although there was a lot of pressure on Twitch from content creators to enable better content moderation (alignment), content creators also used a lot of features that were later incorporated by Twitch into its design (capture). Hence, it seems that in this case, the processes of capture and alignment worked hand-in-hand to bring about design change. More future research is needed to see how these processes interact, reinforce each other, or even work against each other.
Second, in the current analysis, alignment occurs when end-users pressure a platform to change. But other actors—namely governments and advertisers—can also exert pressure on platforms to change (e.g. the European Union's Digital Services Act or the “Adpocalypse” of 2017 on YouTube). Furthermore, at least two other sources of design change are worth noting: in-house innovation (i.e. design initiated or augmented by platforms) and feature removal. These changes fit within the process of adaptive governance, but are much more difficult to identify. Accurately identifying a feature as “original” to a specific platform is difficult. Arguably all online features began with user behavior, suggesting platform change is fundamentally a long process of capture. The removal of features is difficult for a different reason: once gone, there is less of a trace of an old feature, particularly if the feature was not popular in the first place. Nevertheless, a small number of especially (un)popular removed Twitch features can be identified thanks to chroniclers and streamer interviews: the above discussed Soundtrack by Twitch, guest star, ad banner reduction, host mode, and boost train. Motivations for their removal, or their merging with other features, are difficult to trace. However, when a timeline can be constructed, it does much to shed light on platform decision-making.
The resultant theoretical structure—adaptive governance of capture, imitation, and alignment—can make sense of changes on platforms as well as provide a framework for situating platform change as a result of pressure. An effective framework can integrate diverse cases across platforms and over time, enabling researchers to identify pressures and patterns motivating platform design change.
Footnotes
Funding
The authors received no financial support for the research, authorship, and/or publication of this article.
Declaration of conflicting interests
The authors declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.
Data availability statement
Interview data included in this study was collected with the understanding of confidentiality of the participants and non-reproducibility of the full text.
