This article argues that TikTok is more than entertainment; its algorithm—driven by artificial intelligence, automation, and algorithmic governance—wields immense power, raising questions about accountability, transparency, and oversight.
After browsing TikTok for a while, you’ll notice the platform understands your preferences. Videos matching your interests appear instantly, flowing seamlessly. It’s easy to call this “personalization”—a system to help users discover content. But seeing it merely as convenience or efficiency oversimplifies the experience.
What seems like natural recommendations relies on tracking, data analysis, and behavioral prediction. TikTok records signals such as dwell time, scrolling, likes, comments, follows, and replays to infer which content will next capture attention. Platforms don’t just respond to preferences—they shape them. Real-time ranking, filtering, and prioritization determine what is seen, amplified, or buried. Platforms organize the conditions by which attention forms.
This article argues that TikTok shows how platforms use AI, automation, and algorithms to profoundly shape information access and perception—often without users’ awareness. Such systems are not neutral tools but governance mechanisms, raising vital questions about platform responsibility and oversight.
From Personalization to Platform Power
Before examining TikTok’s influence on attention, it is crucial to clarify how artificial intelligence operates on these platforms and why this matters. On digital platforms, AI aligns with data, automation, and platform management. The question is not only what AI can do, but also what it can’t. It is about how it is built, what decisions it supports, and whose interests those decisions serve.
This is where the idea of “algorithmic governance” matters. It shows that artificial intelligence is more than new technology. Instead, it is part of a bigger system that includes platforms, data, automatic decisions, and digital rules. When we think of governance, we often imagine governments, laws, and policies. But platforms shape how people act in ways we do not always notice. They use ranking, prediction, filtering, and tips to determine which information is most likely to get our attention. They also choose which messages we see again and again. As Turow says, personalization is not just about making things work better. It also divides users by how much they are worth to companies, which can create more unfairness in media and shopping ( Turow et al., 2015). So, algorithms shape what we see and also how we focus, interact, and experience social differences.
Why AI Systems Are Not Neutral
AI systems on platforms are never neutral. They are typically designed and optimized around the platform’s own priorities.
Take TikTok as an example:
Its recommendation system is not designed to help users find the most meaningful, balanced, or socially valuable content, but rather to learn from user behavior and push content that is most likely to keep users engaged, clicking, and returning.
Whilst this may appear to be personalization on the surface, it is in fact a business strategy geared towards user engagement. For example, on TikTok, if a user merely pauses briefly, rewatches a video, or engages minimally with a particular type of content, the system swiftly identifies these behaviors as ‘signals of interest’ and continuously reinforces the display of similar content. Whilst the platform appears to be merely responding to user interests, it is in fact constantly amplifying certain behavioral patterns and guiding users to stay within the platform’s feed rhythm.
Crawford’s critique of artificial intelligence is particularly illuminating in this context. She points out that artificial intelligence is not merely a technical tool, but rather an infrastructure, an industrial model, and a means of exercising power. This perspective clarifies why recommendation systems are not harmless, objective, or inevitable tools (Crawford, 2021, p. 18): they are built on the economic logic of large-scale data extraction and behavioral prediction and are thus inevitably embedded within the platform’s commercial interests and value hierarchy. As a result, these systems do not merely reflect users’ existing needs but constantly shape their attention, preferences, and decision-making pathways.
Automation: How Platforms Continuously Operate Behind the Scenes
To better understand the impact of these systems, it is useful to examine how automation consolidates platform power. Platform power remains stable not just because artificial intelligence can predict outcomes. It is also because decision-making is now automated. On digital platforms, many key decisions are no longer made manually or at random. Instead, these decisions are built into systems that run continuously and adjust in real time based on user behavior (AlgorithmWatch, 2019; Thouvenin et al., 2019).
Automation speeds content distribution, but more significantly, it transforms decision execution and transparency. Responses to engagement metrics occur almost instantly, operating in the background. For users, the experience feels natural. Yet it is hard to discern how much content is curated. As Andrejevic (2019) argues, automation embeds human judgments—rules and priorities—into systems, enabling the repeated execution of platform logic.
Algorithmic Governance: When Platform Design Becomes a Form of Control
As we consider how automated decision-making shapes the user experience, it becomes clear that these processes are beginning to assume a governing function. When automated decisions start to shape how users access information and interact with platforms, they become a form of governance. The main question is not just whether platforms “influence” users, but how they set the basic conditions for understanding information, using platforms, and making judgments. This is where the idea of algorithmic governance is important. Governance does not always happen through laws and policies; it can also work through data, platform design, and technical systems (Flew, 2021; Gritsenko & Wood, 2022).
It is important to note that this kind of power rarely appears as direct commands. Platforms usually do not tell users “what you must watch” or “what you should believe.” Instead, they shape daily experiences by building certain rules, priorities, and values into the system. Users might not notice they are being “governed,” but the way they access information, browse content, and what is seen as important is often set by platform design. This is why algorithmic governance is a useful concept. It shows that power in the digital world often works through design, structure, and repetition, not just through obvious control or formal rules.
TikTok as a Case Study of Algorithmic Power
The following example illustrates how continuous scrolling is built into TikTok’s design:
With this framework of algorithmic governance in mind, TikTok can be seen as a prime example of how digital platforms structure attention. TikTok exemplifies governance logic in digital ecosystems, heavily relying on algorithmic recommendation systems. The platform integrates artificial intelligence, automation, and strategic design to orchestrate user attention at scale. TikTok’s influential “For You” page harnesses powerful algorithms to personalize each user’s feed, continually evolving recommendations based on how users engage with content.
TikTok’s recommendation signals include:
- Watch time
- Likes
- Comments
- Follows
- Shares
- Negative feedback signals such as “Not Interested.”
TikTok constantly analyzes these signals to predict what content users want to see next. Instead of sticking to a fixed set of interests, the platform learns from each user’s actions and updates the content it shows.
Building on Just and Latzer’s (2017) concept of “algorithmic selection,” TikTok’s approach to organizing user attention becomes clearer. Instead of simply ranking content, TikTok assigns “relevance” through automated analysis of behavioral data (Just & Latzer, 2017, p. 241). Metrics such as watch time, likes, comments, and swipe patterns serve as both indicators of user preference and criteria for determining subsequent relevance.
Consequently, TikTok personalizes both the content and the very conditions shaping user attention: the platform configures what is perceived as interesting, important, and worthy of ongoing engagement. In this way, users do not simply encounter content that reflects their interests; they move through an information environment that has already been structured for them.
At the same time, this structure remains highly opaque to the average user. Fell and Tan (2025) note that TikTok can push content to groups deemed similar by the platform based on the user characteristics and “personality patterns” it infers. This also helps us understand why, according to the “black box” concept proposed by Pasquale (2015), TikTok’s recommendation system can be viewed as an algorithmic system that is highly opaque to users. Users can see the recommended results, but they cannot truly understand the full predictive and ranking logic behind them. The problem lies not only in this “invisibility,” but also in the fact that this opacity undermines users’ ability to recognize, question, and challenge how the platform organizes their attention.
Therefore, TikTok’s significance lies not merely in its ability to recommend content with precision, but in the way it subtly influences what users notice first, what they continue to follow, and what they perceive as worth watching. Recognizing this influence, users, researchers, and policymakers must actively demand greater transparency, accountability, and oversight. It is crucial to engage critically with these systems to ensure they align with a fairer, more open digital environment.
Who Is Responsible for Algorithmic Influence?
The central issue is no longer the mechanics of TikTok’s content recommendation, but rather the question of accountability when a platform can organize information and attention at such a scale, and the appropriate mechanisms for oversight. Kshetri (2025) argues that TikTok now exemplifies algorithmic power and responsibility, rather than serving solely as a case study of an entertainment platform. When a recommendation system persistently shapes users’ access to information, patterns of attention, and judgment formation, it transitions from a technological tool to an infrastructure warranting rigorous scrutiny and regulation.
What makes the situation even more challenging is that this responsibility is not as easy to define as content liability in traditional media. Platform recommendations do not present the same set of content to everyone; instead, they deliver different content to different users based on user behavior, inferred interests, and the platform’s algorithmic logic. As a result, external researchers, regulators, and even the general public often see only certain controversial outcomes but find it difficult to fully understand how these outcomes were systematically organized step by step.
Precisely because platform visibility is decentralized, regulation becomes even more challenging. Kshetri (2025) emphasizes that the key issue is not merely whether algorithms are “effective,” but rather how platform accountability can be enforced when such algorithmic power operates in a highly personalized, constantly evolving manner that lacks external transparency.
This issue has already made its way onto the regulatory agenda. In its preliminary assessment issued in February 2026, the European Commission identified TikTok’s infinite scroll, autoplay, push notifications, and highly personalized recommendation system as potential risks, concluding that these design features could have a systemic impact on user behavior (European Commission, 2026).
This indicates that the controversy surrounding the platform has moved beyond the realm of academic debate and is gradually shifting toward questions of how to define platform responsibility institutionally, enhance transparency, and establish effective oversight mechanisms.
What truly needs to be scrutinized is not merely whether a particular video is harmful, but whether the platform, through its overall system design, continuously shapes the way people access, interpret, and allocate their attention to information in their daily use. Precisely for this reason, the real issue is not simply whether the platform “influences” users, but whether this influence has, in the absence of transparency and accountability mechanisms, become a form of power that is taken for granted yet difficult to challenge.
The case of TikTok clearly demonstrates that recommendation systems are no longer merely tools to help users find content more efficiently. They are systems that continuously learn from user behavior, automatically assess, and organize the visibility of information. What appears to be a personalized experience is, in fact, a structured one constantly shaped by predictions, rankings, and data-driven decisions. This is why discussions surrounding platforms like TikTok ultimately go beyond user experience or content quality and revolve around power: when systems can influence on a massive scale what people are exposed to, what they follow, and how they interpret information, they are also shaping the public information landscape itself.
This does not imply that technology itself should be dismissed. The central issue is not the utility of these systems, but rather who shapes them, how they influence users, and whether this influence can be understood, questioned, or modified. Internet experiences that seem seamless, natural, and personalized are frequently not spontaneous, but instead result from deliberate orchestration and ongoing optimization. Recognizing this marks the beginning of a more critical inquiry: Given that platforms can shape attention and the information landscape to a significant extent, should they also bear corresponding responsibilities and be held accountable?
Reference
Andrejevic, M. (2019). Automated Media. Routledge. Chapter 3, pp.44-72
Crawford, K. (2021). The Atlas of AI: Power, Politics, and the Planetary Costs of Artificial Intelligence. Yale University Press (pp. 1-21)
European Commission. (2026). Commission preliminarily finds TikTok’s addictive design in breach of the Digital Services Act. European Commission Press Corner. https://ec.europa.eu/commission/presscorner/detail/en/ip_26_312
Flew, T. (2021). Issues of Concern. In T. Flew, Regulating platforms (pp. 79–
86). Polity.
Gritsenko, D., & Wood, M. (2022). Algorithmic governance: A modes of governance approach. Regulation & Governance, 16(1), 45–62. https://doi.org/10.1111/rego.12367
Kshetri, N. (2025). Algorithmic Power and Responsibility: TikTok’s Transition to U.S. Oversight. Computer (Long Beach, Calif.), 58(12), 106–110.
https://doi.org/10.1109/MC.2025.3616484
Pasquale, F. (2015). The black box society : the secret algorithms that control money and information . Harvard University Press.
Turow, J., McGuigan, L., & Maris, E. R. (2015). Making data mining a natural part of life: Physical retailing, customer surveillance and the 21st century social imaginary. European Journal of Cultural Studies, 18(4–5), 464–478. https://doi.org/10.1177/136754941557739
TikTok. (n.d.). How TikTok recommends content. TikTok Support. https://support.tiktok.com/en/using-tiktok/exploring-videos/how-tiktok-recommends-content
TikTok. (2020). How TikTok recommends videos #ForYou. TikTok Newsroom. https://newsroom.tiktok.com/how-tiktok-recommends-videos-for-you?lang=en
Be the first to comment