Transparency Discourse on Digital Platforms: A Comparative Textual Analysis of Platform Reports and Regulatory Texts in the EU and Türkiye

Lectio Socialis RE SEARCH ARTICLE 2026: 1921434 DOI: 10.47478/lectio.1921434 CORRESPONDING AUTHOR Emel Dikba ş Torun, Pamukkale University, Faculty of Communication, 20160, Denizli, TÜRK İ YE . Email: etorun@pau.edu.tr The Author(s). Published by Lectio Socialis . This is an Open Access article distributed under the terms of the Creative Commons Attribution - 4.0 License (CC BY 4.0) , which permits re - use, distribution, and reproduction in any medium, provided the original work is properly cited . Transparency discourse on digital platforms: A comparative textual analysis of platform reports and regulatory texts in the EU and Türkiye Emel Dikba ş Torun a a Pamukkale University, Faculty of Communication, Denizli, TÜRK İ YE. ARTICLE HISTORY Receiv ed: 02.04.2026 Accepted: 09.07.2026 Published: 16.07.2026 KEYWORDS digital platforms, transparency reports, algorithmic governance, Digital Services Act (DSA), Turkish regulatory framework (Law No. 7253) Introduction Digital platforms form the core infrastructure of today's information society, shaping billions of users' access to information, freedom of expression, and public debate. Multinational platforms such as Meta, X (formerly Twitter), YouTube, and TikTok autom ate content moderation decisions through algorithmic systems and present these processes to the public through transparency reports. However, whether these reports truly ensure accountability is central to academic and regulatory debate (Gillespie, 2018; G orwa, 2019). The platforms' discourse on transparency may function not only as a response to regulatory pressures but also as a way to mask the complexity of algorithmic decision - making processes (Ananny & Crawford, 2018). ABSTRACT This study examines transparency reporting in digital platform governance through a comparative analysis of platform reports, the European Union’s Digital Services Act (DSA), and Türkiye’s Law No. 7253. Drawing on surveillance capitalism, disciplinary powe r, and critical platform studies, the research employs systematic qualitative content analysis using MAXQDA software. The analysis covers 65 transparency reports and two regulatory texts published by Meta, X (formerly Twitter), YouTube, and TikTok between 2017 and 2025. Inter - coder reliability was determined as κ = 0.82 (Cohen’s Kappa), indicating strong agreement. The findings show that platforms construct an illusion of algorithmic neutrality through a demonstrative and metric - based discourse, framing aut omated systems as neutral and objective while leaving moderation processes insufficiently explained. Comparative analysis indicates that EU regulatory texts place greater emphasis on accountability, while Turkish regulatory texts focus more on content remo val and procedural compliance. The study argues that transparency reports primarily function as tools for demonstrating accountability and regulatory compliance rather than as mechanisms of democratic oversight. It contributes to debates on platform govern ance and regulatory accountability across different regulatory contexts. The findings further suggest that transparency is often limited to measurable metrics. This allows platforms to report activity without meaningfully explaining how moderation processe s function. Torun 2 The European Union Digital Services Act (DSA), which came into force in 2022, imposed comprehensive transparency obligations on very large online platforms (VLOPs) (European Union, 2022). The DSA mandates that platforms regularly report on their content mo deration decisions, algorithmic systems, and risk assessments. In Türkiye, Law No. 7253, which came into force in 2020, obligates social network providers to appoint local representatives, respond to content removal requests, and submit regular reports (BT K, 2020). While both regulations aim to increase platform transparency, their approaches and scopes differ significantly. In recent years, the transparency database introduced by the DSA has allowed researchers to systematically examine platform moderation practices. Kaushal et al. (2024) analyzed the potential of the DSA transparency database to provide automated transparenc y from a legal and empirical perspective, highlighting the structural limitations of the database. Trujillo et al. (2023) questioned the usability of the DSA database in monitoring self - reported moderation actions by platforms, drawing attention to the rel iability issues of self - reporting systems. Drolsbach and Pröllochs (2024) showed that the analysis of content moderation practices in the EU through the DSA database yielded significant findings, but data quality and consistency issues persisted. The literature points to the need to move beyond transparency reports towards independent auditing and platform observability (Leerssen, 2024; Terzis et al., 2024). Current platform transparency strategies are limited to demonstrating regulatory compliance, and algorithmic opacity is a structural feature of these strategies. Shahi et al. (2025) analyzed one year of data from the DSA transparency database to reveal wha t the database shows and does not show regarding platform auditing. Piovano (2025) discussed how transparency compliance can be made more effective by proposing a socio - technical framework for automating the DSA application. Strowel (2024), in his study, e valuated the potential of DSA to use transparency as an effective tool in preventing the spread of disinformation. Accuracy and accountability of content moderation are among the core objectives of DSA. Wei et al. (2024) examined the operationalization of content moderation accuracy within the scope of DSA, questioning the reliability of accuracy metrics reported by th e platform. Roberts (2019) addressed the invisible dimensions of content moderation, revealing the hidden aspects of moderation efforts. In his research, Mittelstadt (2016) emphasized the importance of auditing for transparency in content personalization s ystems and discussed how accountability of algorithmic systems can be ensured. Accountability of automated decision - making systems stands out as a critical dimension of platform governance. Cobbe et al. (2021) emphasized the importance of reviewable automated decision - making, arguing that algorithmic systems should be open to human o versight. Bucher (2018) examined the political power and policies of algorithms, analyzing the societal consequences of "if...then" logic. Napoli (2019) addressed the public benefit dimension of social media, discussing the social responsibilities of platforms. Suzo r (2019) criticized the lack of transparency in platform rules, examining the hidden rules governing our digital lives. The challenges and information overload of platform regulation limit regulatory effectiveness. Lodder and Carvalho (2022) state that online platforms need new requirements regarding moderation, ranking, and traceability, leading to an information tsunami. Ilieva (2023) assessed the effectiveness of regulatory tools by examining the role of EU platfo rm Lectio Socialis 3 regulations in combating online disinformation. Quintas - Froufe et al. (2024) questioned the adequacy of corporate policies by analyzing TikTok's policies to protect young audiences from disinformation. Platforms ’ strategic responses are shaped by regulatory pressures. Islam (2025) examined Meta's algorithmic transparency and competitive advantage strategy against European data privacy and digital services regulations, investigating how it balanced regulatory compl iance with commercial interests. Mishra and Agrawal (2025) investigated how the principles of transparency and accountability are implemented in DSA, revealing the strengths and weaknesses of the regulatory framework. This study examines the transparency reports of platforms and relevant regulatory texts using systematic qualitative content analysis. While previous studies have largely focused on disclosure practices and compliance requirements, the present study focuse s on how transparency and accountability are presented across platform reports and regulatory texts. To address this aim, the following research questions were developed: RQ1: What discursive strategies are predominantly used in the transparency reports of Meta, X, YouTube, and TikTok, and how do these strategies differ among the platforms? RQ2: What discursive priorities regarding transparency and accountability do the DSA and TR 7253 regulatory texts reflect, and how do these priorities align with or diverge from the platform reports? RQ3: To what extent do transparency claims contribute to the visibility of platform governance practices, and how are these claims associated with procedural compliance? This study, building upon Leone de Castris's (2025) classification of platform transparency types, utilizes Han's (2015) critique of the transparency society, Foucault's (1977) theory of discipline and surveillance, and Zuboff's (2019) concept of surveilla nce capitalism as its theoretical framework. The methodology section details the systematic coding process using Grimmer and Stewart's (2013) text analysis approach. Theoretical framework Critique of the transparency society and surveillance The contemporary demand for transparency in digital platforms must be understood against the backdrop of broader social and political transformations. Byung - Chul Han (2015) argues that contemporary society is besieged by the ideology of transparency, which , while presented as a liberating tool, actually creates a new form of control. According to Han, the transparency society eliminates privacy by demanding that everything be made visible, keeping individuals under constant surveillance. Transparency report s from digital platforms provide a concrete example of this paradox. Platforms claim to be transparent by making content moderation decisions visible through numerical data, but they conceal the essence of their algorithmic decision - making processes. Foucault's (1977) theory of discipline and surveillance offers a fundamental framework for understanding how power operates in modern society. Using the Panopticon metaphor, Foucault demonstrated that surveillance operates not only in physical spaces but a lso at the level of information production and discourse. Digital platforms function as a contemporary Panopticon, monitoring and recording every movement of users. Transparency reports are used as a tool to legitimize this surveillance process. By claimin g to be transparent while monitoring users, the platforms establish an asymmetrical power relationship. Torun 4 Ananny and Crawford (2018) critically examine the limitations of the transparency ideal in the context of algorithmic accountability. Using the concept of “seeing without knowing,” the authors demonstrate that platform transparency reports offer superficia l information but are insufficient for understanding how algorithmic systems work. Transparency should not be limited to simply disclosing information; it must also ensure that this information is understandable and actionable. Platform transparency narrat ives are often filled with technical jargon and complex statistics, making them difficult for ordinary users to comprehend. Gillespie (2018) notes that platforms assume the role of “guardians of the internet” in content moderation, but this role is problematic in terms of transparency and accountability. Platforms automate decisions about which content to remove, which accounts to suspend, and which content to promote using algorithmic systems, but avoid explaining the logic behind these decisions. Gillespie characterizes the moderation decisions of platforms as “secret decisions” and emphasizes that these decisions are problema tic in terms of democratic values. Surveillance capitalism and behavioral data extraction Shoshana Zuboff (2019), using the concept of surveillance capitalism, suggests that the business models of digital platforms are built upon the extraction and commercialization of user data. According to Zuboff, surveillance capitalism uses big data analyt ics and machine learning techniques to predict and direct user behavior. Platform transparency reports conceal this data extraction process, focusing only on visible activities such as content moderation. However, the platforms' true power stems from gener ating advertising revenue by collecting user behavioral data. Zuboff identifies four fundamental stages of surveillance capitalism: data collection, data analysis, behavioral prediction, and behavioral intervention. Digital platforms create massive datasets by recording every click, like, share, and even screen time of users. This data is analyzed using machine learning algorithms to predict users' future behavior. These predicted behaviors are then manipulated through targeted advertising and content recommendations. Transparency reports only reveal a small part of t his process (content moderation), concealing the actual economic value creation process. Parsons (2019) examines the ineffectiveness of voluntarily generated transparency reports, demonstrating that platforms strategically use these reports to mitigate regulatory pressure. Parsons notes that transparency reports often offer selective informati on, with platforms highlighting data that presents them in a positive light. For example, platforms emphasize the number of removed pieces of content while avoiding information about content that was mistakenly removed or harmful content that was not remov ed. Suzor (2019) criticizes the lack of transparency in platform rules by examining the hidden rules that govern our digital lives. Suzor notes that platform terms of use and community standards are often vague and open to interpretation, making it difficult t o protect users' rights. Transparency reports show how these rules are implemented but do not explain how the rules themselves are determined and changed. Types of platform transparency and observability Leone de Castris (2025), by systematically analyzing discourses on platform transparency, identifies three basic types of transparency: procedural transparency, algorithmic transparency, and organizational transparency. Procedural transparency refers to pl atforms Lectio Socialis 5 disclosing their content moderation processes and statistics. Algorithmic transparency involves explaining how algorithms work and what criteria they use. Organizational transparency encompasses disclosing the organizational structure, decision - making proc esses, and stakeholder relationships of platforms. Leone de Castris indicates that platforms mostly focus on procedural transparency but fall short in algorithmic and organizational transparency. Urman and Makhortykh (2023) compared transparency reporting practices across major online platforms and reported important differences in reporting structures, disclosure practices, and transparency priorities. Reid and Ringel (2025) argued that transparency reports increasingly function as public - facing governance documents through which platforms communicate their moderation and accountability practices to regulators and users. Leerssen (2024) argues for the need to move beyond the “black box” metaphor, advocating for a shift from algorithmic transparency to platform observability. Leerssen emphasizes that transparency should not be limited to disclosing information, but should a lso include providing tools that allow independent researchers and regulators to audit platforms. Platform observability includes tools such as API access, data sharing, and independent auditing. Fung et al. (2007) comprehensively examine the promises and dangers of transparency, demonstrating that full disclosure policies do not always yield positive results. The authors emphasize that for transparency to be effective, information must be accurate , timely, understandable, and actionable. Transparency reports from platforms often fail to meet these criteria. Information is presented in technical jargon, published with delays, and is difficult for users to act upon. Gorwa (2019) defines platform governance and examines how platforms assume governance functions in areas such as content moderation, user behavior regulation, and algorithmic decision - making. Gorwa notes that platforms, as private companies, assuming roles similar to public authorities creates problems of democratic legitimacy. Transparency and accountability in platform governance are critical to mitigating this legitimacy problem. Platform governance and political science Viewing digital platforms solely as technological infrastructure overlooks their political dimension. Gorwa (2019) argues that platforms effectively assume functions similar to public authorities in areas such as content moderation and algorithmic decision - making, raising serious questions about democratic legitimacy. Platforms become governance actors, acting as rule - makers, interpreters, and enforcers. Political science literature addresses this transformation within the framework of sovereignty debates. Frosio & Geiger (2023) indicates that the EU used the discourse of “taking back control” while legitimizing the DSA, and that the law is positioned as a tool for asserting its claim to digital sovereignty. In this context, the DSA is not merely a regulatory text, but an attempt by the EU to establish political authority against US - based platform companies. Samuelson and Helberger (2024), on the other hand , interpret the DSA as the beginning of a global transparency regime and discuss the potential for exporting regulatory norms to other states. The concept of algorithmic governance captures these power relations at the technical level. Niva Elkin - Koren and Maayan Perel (2019) demonstrate that platform actors employ algorithmic enforcement mechanisms. Their study reveals that enforcement decisions Torun 6 operate simultaneously within both private and public legal domains, and that this overlap generates significant accountability gaps. Francesca Di Porto and Ginevra Zuppetta (2021) further argue that algorithmic disclosure obligations can be institutionali zed through hybrid public – private regulatory frameworks. Husovec (2024) emphasizes that the success of DSA depends not only on the legal text but also on the strength of implementation networks and civil society monitoring capacity, while Griffin (2025) wa rns that the systemic risk management approach risks depoliticizing distributive conflicts and delegating responsibility to platforms. This tension defines a political space in which transparency reports are directly involved: reports become a tool for man aging regulatory pressure, thus evolving from technical documents to political documents. Although Han (2015), Foucault (1977), Zuboff (2019), and Gorwa (2019) draw upon different theoretical traditions, they converge on a single point: visibility is never neutral, and making something visible is not an elimination of power but an exercise of i t. Increased transparency does not automatically guarantee accountability; visibility itself can transform into a mechanism through which power is exercised, selectively managed, and legitimized (Ananny & Crawford, 2018; Flyverbom, 2019). Where these four approaches diverge is in the nature of the power inherent in this visibility. For Foucault (1977), visibility is a disciplinary technique enacted through institutional arrangements and operating via hierarchical observation and normalizing judgment. In con trast, for Han (2015), visibility is a form of power that operates through a process of voluntary self - disclosure under conditions of neoliberal subjectivity, requiring no external coercion. Zuboff (2019) shifts this dynamic to the political economy of dat a extraction, where the visibility of user behaviors rather than that of institutions is transformed into raw material for profit. Finally, Gorwa (2019) treats visibility as an issue of corporate governance by questioning who has the authority to oversee a nd regulate platform behavior. These perspectives therefore operate at complementary levels of analysis and together provide the theoretical foundation for this study. Han (2015) provides a philosophical - historical perspective on transparency, Foucault (1977) a structural - theoretical ac count of visibility and discipline, Zuboff (2019) a political - economic explanation of behavioral data extraction, and Gorwa (2019) an institutional perspective on platform governance. When these perspectives are considered together, it becomes evident that platform transparency reports actually serve multiple functions. These functions can be described as supporting organizational legitimacy, facilitating procedural disclosure, and creating forms of strategic visibility. On the other hand, it appears that t hese reports do not sufficiently explain key aspects of algorithmic decision - making processes. This multi - level perspective provides the analytical framework for interpreting the empirical findings presented in the following sections. Methodology Qualitative research design, data collection and analysis process This study was conducted using a systematic qualitative content analysis method because it enables the systematic examination of recurring patterns, themes, and discursive structures across large collections of textual documents (Mayring, 2014). The resear ch examined a total of 65 transparency reports published by Meta, X, YouTube, and TikTok platforms between 2017 and 2025, as well as two regulatory documents: the European Union Digital Services Act (DSA) and Turkey's Law No. 7253. Platform reports were co llected from Lectio Socialis 7 each platform's official transparency websites. Meta Platforms (2017 – 2025) published 16 reports with a total of 12,386 words, X Corp (2018 – 2024) published 13 reports with a total of 46,044 words, Google LLC (2019 – 2025) published 24 reports with a total of 2,355 words, and TikTok (2020 – 2025) published 12 reports with a total of 1,464 words. All relevant published reports were included in the research in their entirety. The reports also varied considerably in length and format. While some platforms published detailed narrative reports, others relied more heavily on dashboard - style reporting. To reduce the effect of these differences, normalized frequencies were calculated and used throughout the analysis. The analysis followed the principles of qualitative content analysis proposed by Mayring (2014), in which categories are systematically developed and applied to identify recurring patterns within textual data. MAXQDA was used to organize the coding process , manage the documents, and compare the frequencies and relationships among the identified categories. This procedure is consistent with the qualitative text analysis approach described by Kuckartz (2014). Reports are tagged with metadata such as publication date, period covered, word count, and content type. The DSA text with a total of 82,719 words was taken from the Official Journal of the European Union (European Union, 2022). The text of Law No. 7253 wi th a total of 1,429 words was obtained from the official website of the Information and Communication Technologies Authority (BTK) of Republic of Turkiye. These word counts were used as the basis for normalized frequency calculations. In their study discussing the use of text analysis methods in political texts, Grimmer and Stewart (2013) suggest using multi - coder reliability protocols to increase the relia bility of systematic coding processes. The data collection process involved systematically scanning and indexing the transparency reports of each platform. Transparency reports published by Meta, X (Twitter), YouTube, and TikTok in different years were systematically reviewed, and relevant repo rts were accessed through the platforms' official websites and archives. Due to the reports being presented in different formats (PDF, web - based reports, etc.) and the lack of a standard structure, the data collection process was carried out carefully and in multiple stages. To parse and analyze the text content of PDF reports, the pdfplumber library, which runs in Python, was used. During this process, reports were processed at the page level, text was cleaned, and the data was converted into a suitable structure for analysis . Since there were terminological and structural differences between reports from different platforms, manual checking and editing were performed during the dataset creation process to allow for content comparison. Special attention was paid to maintaining data integrity and ensuring analyzability. Thus, an integrated and consistent dataset was obtained for use in the analysis process. Some reports also contained visual elements such as charts and dashboards. However, the focus of this study was the textual construction of transparency and accountability. For this reason, visual elements were not coded separately and were considered only as supporting components of the reports. A thematic analysis approach was adopted in the analysis of the obtained qualitative data. The data analysis process was carried out using MAXQDA software, and the texts were examined in detail and manually coded by the researcher using a keyword - based sys tematic coding approach. The codes were grouped according to content similarities, and themes were developed. A data - driven (inductive) and theoretically consistent (deductive) approach was used in the creation of the themes (See Figure 1). Torun 8 Figure 1. Coding process in MAXQDA and the code – corpus tree. In the initial stages of the coding process, given the broad scope of the dataset, the code and themes were reviewed in order to avoid overlooking potential patterns. All coding decisions were made by the researcher based on the context of the data. The co ding scheme was developed and refined through pilot coding, and the entire coding, interpretation, and theme generation process was carried out manually by the researcher. Coding categories were developed within the framework of Leone de Castris's (2025) classification of platform transparency types, Han's (2015) critique of the transparency society, and Zuboff's (2019) concept of surveillance capitalism. A total of 10 code categories and keyword lists were created for each code category (see Table 1). The keyword lists were tested during the pilot coding process, and necessary corrections were made. The Turkish regulatory text was coded in its original form by a Turkish - nati ve researcher, with advanced proficiency in English, enabling direct application of coding categories in both languages without translation, based on the conceptual meaning of the categories. The coding process was carried out in three stages: pilot coding , revision of category definitions, and full corpus coding. Ambiguous cases were reviewed within their textual context before final coding decisions were made. The coding framework presented in Table 1 illustrates the analytical categories used throughout the study. These categories were applied consistently across all platform reports and regulatory documents to identify recurring patterns in transparency discou rse. The coding process focused on the contextual meaning of each text segment rather than the frequency of individual words, allowing broader discursive themes to emerge from the data. The full coding scheme, comprising the category definitions and keywor d lists, together with the normalized frequency dataset underlying Figures 3 – 7, is openly available Lectio Socialis 9 in the Lectio Socialis Zenodo community (Dikba ş Torun, 2026; see Data Availability Statement). Table 1. Coding categories, definitions, and keywords # Code Category Definition Keywords 1 Quantified Transparency Transparency discourse conveyed through statistical data, percentages, numbers, and metrics. percent, %, rate, count, metric, transparency report, statistic, total, volume, figures, indicators, benchmark, dataset, reporting period, quarterly, annually, trend, increase / decrease, proportion, ratio 2 Algorithmic Neutrality Claim The claim that algorithms are objective, impartial, and fair. algorithm, AI, automated detection, fair, neutral, machine learning, algorithmic, AI, automated detection, fair, neutral, machine learning, objective, unbiased, non - discriminatory, consistent, accuracy, precision, automated decision - making, model performan ce, detection system, integrity 3 Performative Compliance Discourse of demonstrating adherence to regulatory requirements. comply, compliance, policy, regulation, commitment, standard, guideline, comply, compliance, policy, regulation, commitment, standard, guideline, obligation, adherence, enforcement framework, compliance mechanism, governance, regulatory alignment, implemen tation, accountability, transparency commitment 4 Political Subjectivity Emphasis on policies, rules, and community standards. community standards, policy, rule, terms of service, content policy, community standards, policy, rule, terms of service, content policy, content moderation, editorial decision, platform position, discretion, judgment, prioritization, policy interpretation , sensitive content, harmful content classification 5 Automated Enforcement Discourse of automated content removal and account suspension. automated removal, proactive rate, account suspension, automated removal, proactive rate, bandwidth throttling, account suspension, detection rate, proactive detection, automated flagging, removal rate, system intervention, content filtering, AI enforcemen t, auto - detection, scalability 6 Government Request Compliance Fulfillment of requests from government and law enforcement agencies. government request, law enforcement, legal request, court order, government request, law enforcement, legal request, court order, legal obligation, authority request, regulatory request, compliance rate, data disclosure, judicial order, national authority, law enforcement request, transparency to authorities 7 Behavioral Data Extraction Collection and use of user data. behavioral data, targeted advertising, profiling, data collection, user data, behavioral data, targeted advertising, profiling, data collection, user data, tracking, analytics, user behavior, engagement data, personalization, recommendation system, ad targ eting, profiling system, data processing, metadata 8 User Agency Limitation Restriction of users' control over content and accounts. account termination, content removal, appeal, user control, restriction, account termination, content removal, appeal, user control, restriction, suspension, limitation, lack of control, appeal process, user rights, access limitation, content visibility re striction, enforcement action 9 Regulatory Asymmetry Regulatory obligations and asymmetric power relations. DSA, GDPR, regulation, asymmetry, compliance obligation, regulatory, DSA, GDPR, regulation, asymmetry, compliance obligation, regulatory, obligation, compliance burden, regulatory pressure, legal asymmetry, power imbalance, jurisdiction, cross - border regul ation, enforcement disparity, regulatory framework 10 Selective Visibility Withholding or non - disclosure of certain information. trade secret, confidential, not disclosed, selective, proprietary, undisclosed, trade secret, confidential, not disclosed, selective, proprietary, undisclosed, disclosure limitation, partial transparency, restricted information, withheld data, redacted, no n - reporting, opacity, limited disclosure, unavailable data Torun 10 Inter - coder reliability was assessed using Cohen's Kappa. A second coder independently coded a randomly selected 20% of the dataset using the same coding scheme. The resulting Cohen’s Kappa value was 0.82, indicating strong agreement. Prior to this stage, the coding scheme was clarified through pilot coding to ensure consistency in the interpretation of categories. This value indicates a high level of inter - coder agreement (Landis & Koch, 1977). The keyword lists were used as guiding indicators rather than strict inclusion criteria, and all coding decisions were made based on contextual interpretation. Normalized frequency values were calculated by dividing the raw frequency in each category by the total number of words and multiplying by 1,000. The generated code structure was evaluated by a domain expert, and necessary adjustments were made to finalize it. Although keywords were used as a guide in the coding process, coding was not based on automatic keyword counting. Each instance was carefully examined within its context to determine whether it reflected the intended meaning of the category. In other words , the presence of a keyword alone was not considered sufficient for coding unless it was supported by the broader textual context. This approach allowed the analysis to remain interpretive and ensured that coding decisions were grounded in meaning rather t han frequency alone. The comprehensive analysis process of the research is given in Figure 2. Figure 2. Qualitative analysis process flowchart (showing data collection, preprocessing, coding, normalization, and interpretation stages). Lectio Socialis 11 The data obtained from the coding process were exported after the necessary calculations were completed in MAXQDA, and then redrawn using Python - based visualization tools to more clearly reveal comparative patterns. In this context, heat maps were created to show the distribution of codes within the data sources, stacked bar graphs were used to present the comparative structure of code frequencies, and radar graphs were used to visualize the relationships between themes. The graphs obtained during the visualization process were structured to directly reflect the data structure, and the interpretation pro cess was carried out under the supervision of the researcher. Findings General distribution and platform profiles The analysis results show significant differences in the transparency discourse of the platforms. Normalized frequencies, heat map and the profiles of the platforms in the code categories are shown in Figures 3 and 4, respectively. Figure 3. Normalized frequencies heat map (dark colors represent high frequencies, light colors represent low frequencies): platform transparency reports and regulatory documents (per 1,000 words). In the category of Algorithmic Neutrality Claims, Meta leads with a frequency of 23.58. The text DSA shows a frequency of 15.10, X 6.28, YouTube 11.04, and TikTok 10.93. The text TR 7253 shows a frequency of only 0.70, almost completely omitting any mentio n of algorithmic neutrality. These findings indicate that Meta heavily utilizes a strategy of emphasizing the neutrality of artificial intelligence and algorithms in its transparency discourse. In the Automated Enforcement category, YouTube stands out with a frequency of 23.78 and X with 17.72. TikTok shows a frequency of 3.42, Meta 1.78, DSA 0.73, and TR 7253 0.00. The fact that YouTube and X heavily use the rhetoric of automated content removal and account suspension indicates that these platforms' moderation strategies are based on automation. Torun 12 In the category of Political Subjectivity, TikTok has the highest frequency with 47.13 and YouTube with 45.44. X shows a frequency of 15.96, Meta 14.21, DSA 3.16, and TR 7253 0.00. TikTok and YouTube's tendency to emphasize policies, rules, and community s tandards indicates that these platforms have adopted a rule - based approach to content moderation. In the category of Government Request Compliance, TikTok shows a frequency of 8.88, Meta 6.14, YouTube 3.82, X 2.95, DSA 0.18, and TR 7253 0.70. TikTok's tendency to emphasize government requirements indicates that the platform has adopted a compliant disc ourse in the face of regulatory pressures. In the Behavioral Data Extraction category, Meta shows a frequency of 5.89, TikTok 2.73, DSA 2.42, X 0.74, and YouTube with TR 7253 0.00. Meta's tendency to emphasize user data reflects that the platform's business model is based on data extraction. In the category of Regulatory Asymmetry, DSA stands out with a frequency of 10.31 and TR 7253 with 6.30. Meta shows a frequency of 3.55, while X, YouTube, and TikTok show a frequency of 0.00. DSA's tendency to emphasize regulatory obligations and asymmetri c power relations indicates that the regulatory text aims to ensure accountability in platform governance. In the Quantified Transparency category, YouTube is the clear leader with a frequency of 92.99. X follows with 79.47, TikTok with 49.18, and Meta with 45.37. Regulatory texts exhibit significantly lower values in this category, with TR 7253 ranking at 16.1 0 and DSA at 10.07. YouTube and X's strong emphasis on quantitative data presentation reveals that these platforms build their understanding of transparency on quantitative reporting practices. In contrast, the low quantified transparency frequencies of DS A and TR 7253 indicate that regulatory texts are structured more around normative obligations than quantitative accountability. In the Performative Compliance category, TikTok has the highest value with a frequency of 8.20, while X is second with 4.06. DSA shows a frequency of 1.81, YouTube 1.27, Meta 0.57, and TR 7253 0.00. TikTok's more intensive use of performative compliance di scourse compared to other platforms indicates that it has adopted a strategy of projecting an image of compliance, particularly in the face of Western regulatory pressures. The fact that TR 7253 shows zero frequency in this category reflects the national r egulatory text's preference for a framework of direct sanctions and obligations rather than compliance performance. In the User Agency Limitation category, Meta shows a frequency of 0.40 and X shows 0.74. DSA exhibits a relatively low value in this category with a frequency of 0.34. YouTube, TikTok, and TR 7253 have no mention in this category with a frequency of 0.00. Meta and X's discourse restricting user agency — account restrictions, content access limitations, and platform authority over user control — is more prominent compared to other platforms. DSA's modest frequency in this category can be considered an indication that the regulatory text's framework for protecting user rights remains limited at the discursive level. In the Selective Visibility category, TikTok is the only platform with a significant value at a frequency of 0.68. Meta shows a frequency of 0.08, DSA 0.45, while X, YouTube, and TR 7253 exhibit a frequency of 0.00 in this category. The Selective visibilit y has the lowest overall frequency among all categories; this indicates that platforms and regulatory texts avoid explicitly addressing their practices of selectively managing content visibility in transparency reports. TikTok's relatively high value in th is category suggests that the Lectio Socialis 13 platform ’ s recommendation algorithms and content boosting/pushing mechanisms are partially accepted at the discursive level. The grouped bar graph presented in Figure 4 shows the cross - platform frequency differences for each coding category in more detail. Figure 4 also shows that the categories are not distributed evenly across platforms. While some themes receive greater empha sis in particular reports, others appear less frequently. In particular, Quantified Transparency is consistently the most prominent category across platform reports, whereas Political Subjectivity and Automated Enforcement receive greater emphasis in the regulatory texts. This suggests that platforms primarily communicate transparency through measurable reporting practices, while the regulatory documents place greater emphasis on governance responsibilities and legal obligations. Figure 4. Grouped bar chart: normalized coding frequencies by platform and category (per 1,000 words). Figure 5 shows clear differences in the distribution of transparency categories across platforms and regulatory texts. TR 7253 is dominated by Quantified Transparency (67.6%), representing the most one - dimensional discursive profile among all documents. By contrast, the DSA exhibits the most balanced distribution, with Algorithmic Neutrality Claim (33.9%), Regulatory Asymmetry (23.1%), and Quantified Transparency (22.6%) emerging as its dominant categories. Among the platforms, X devotes 62.1% of its discou rse to Quantified Transparency, whereas TikTok presents the most balanced profile, with Quantified Transparency (37.5%) and Political Subjectivity (35.9%) receiving comparable emphasis. These differences suggest that platforms and regulatory texts construc t transparency through distinct discursive priorities. 67.6% of TR 7253's total discourse, consists of the Quantified Transparency category; this is the most one - dimensional profile among all documents. DSA, on the other hand, exhibits the most diverse distribution: quantified transparency stands out at 22.6%, the claim of algorithmic neutrality at 33.9%, and regulatory asymmetry at 23.1%. Among social media platforms, X dedicates 62.1% of its discourse to Quantified Transparency, while TikTok shows the most balanced distribution, with quantified transparency at 37.5% and political subjectivity at 35.9%. Torun 14 Figure 5. Stacked bar chart: composite of transparency discourse by coding category, showing percentage distribution per platform. Illustrative coding examples from transparency reports To provide concrete examples of how these discursive patterns appear in practice, selected excerpts from platform transparency reports are presented below. “Because participants’ flags have a higher action rate than the average user, we prioritize them for review.” (Selective Visibility) This statement (YouTube, Selective Visibility) speaks to RQ1. By prioritizing certain flaggers without disclosing the criteria for that prioritization, the platform enacts transparency while simultaneously rendering its decision - making logic proprietary. T his exemplifies Ananny and Crawford's (2018) “seeing without knowing” pattern, in which visibility of outcomes coexists with opacity of the processes that produce them. “Using a combination of people and technology, we remove comments that violate our Community Guidelines.” (Automated enforcement) This statement (YouTube) presents moderation as a technical and partly automated process, yet it does not explain how decisions are made or how responsibility is distributed between human reviewers and automated systems. “We rely on teams around the world to review flagged content and remove content that violates our Community Guidelines.” (Performative Compliance) This statement (YouTube) emphasizes global scale and human review but does not provide information about how consistency or decision criteria are ensured. These examples suggest how transparency is primarily presented through descriptions of moderation act ivities and reporting procedures. “We received a blocking access decision for 2 Instagram posts alleging that an investigation had been launched into Çaycuma’s Chief Prosecutor, who reportedly had his girlfriend's ex - boyfriend detained on charges of ‘illegally obtaining personal data’ for opening Instagram accounts in his name. The blocking access decision required that the decision must be implemented immediately, at the latest Lectio Socialis 15 within 4 hours of notification, and the risks of not complying include potential civil and criminal liability.” (Government Request Compliance) This statement (Meta, Government Request Compliance) speaks to RQ2. By attributing content removal to a court order rather than to platform policy, Meta frames moderation as a response to external legal authority instead of an internally governed decision. This reflects Gorwa's (2019) argument that platform governance is shaped through the interaction between corporate decision - making and external regulatory institutions. “We may withhold content in a specific country if we receive a valid legal request that complies with local laws.” (Government Request Compliance) This statement from X’s Transparency Report shows that content moderation decisions are shaped by national legal frameworks. Platform actions depend not only on internal policies but also on external governmental requests. “When we receive a request, we review it for compliance with applicable laws and our policies before taking action.” (Performative Compliance) This statement (X, Performative Compliance) speaks to RQ3. By presenting moderation as a procedural review while leaving the criteria and evaluation process unexplained, the platform emphasizes compliance without making its decision - making practices fully transparent. This reflects Ananny and Crawford's (2018) argument that transparency may disclose procedures without necessarily making them understandable. “We are committed to maintaining a safe and welcoming environment by proactively identifying and removing content that violates our Community Guidelines.” (Performative Compliance) This statement from TikTok’s transparency report frames moderation as a proactive and protective process. However, it does not explain how content is identified, what criteria are applied, or how errors are handled. “We use automated systems to detect and remove violating content at scale.” (Automated Enforcement) This statement (TikTok, Automated Enforcement) speaks to RQ1. By emphasizing automated detection and removal at scale, the report presents moderation primarily as a technical process while providing little information about how automated decisions are eval uated or verified. Within the coding framework, this excerpt was classified as Automated Enforcement because the emphasis is placed on automated intervention rather than on the decision - making process itself. “Providers of intermediary services shall include information on any policies, procedures, measures and tools used for the purpose of content moderation, including algorithmic decision - making and human review…” (Regulatory Asymmetry) This statement from EU DSA Regulation requires platforms to disclose their content moderation processes, including algorithmic systems. It reflects a regulatory expectation of transparency at the level of procedures and decision - making structures. “Online platforms should ensure that recipients of their service are appropriately informed about how recommender systems impact the way information is displayed…” Torun 16 (Selective Visibility) This statement (EU DSA) highlights the obligation to explain how algorithmic systems shape information visibility. It shows that transparency is expected not only in moderation but also in algorithmic curation processes. “Providers of very large online platforms should assess how their algorithmic systems contribute to systemic risks…” (Regulatory Asymmetry) This statement (EU DSA, Regulatory Asymmetry) speaks to RQ2. By requiring platforms to disclose information about algorithmic decision - making and content moderation procedures, the DSA establishes a broader regulatory framework for transparency than curren t platform reporting practices. This finding is consistent with Bradford's (2020) argument that the DSA functions as a standard - setting regulatory framework extending beyond the EU. These examples show that the DSA defines transparency as a multi - dimensional obligation that extends beyond reporting to include algorithmic explanation and risk assessment. These examples indicate that the DSA approaches transparency as a broader accounta bility mechanism. “Social network providers are obliged to respond to content removal requests within 48 hours.” (Performative Compliance) Above statement from Turkish Law 7253 Regulatory frames platform responsibility primarily as a time - bound obligation to respond to removal requests. It focuses on procedural compliance rather than the transparency of decision - making processes. “Social network providers shall take necessary measures to remove content or restrict access in accordance with legal requests.” (Government Request Compliance) This statement (TR 7253) emphasizes content removal and access restriction as the core regulatory focus. It does not include requirements related to algorithmic transparency or user rights. Unlike the DSA, which requires platforms to explain algorithmic pr ocesses and assess systemic risks, Law No. 7253 primarily focuses on content removal obligations and procedural compliance. In contrast, Law No. 7253 places greater emphasis on procedural obligations and content - related responsibilities. Meta: Algorithmic neutrality claim and behavioral data extraction discourse Meta's transparency reports heavily utilize the claim of algorithmic neutrality (23.58). Meta emphasizes that artificial intelligence and machine learning systems make objective and fair decisions in content moderation. Terms such as "AI," "feed," and "alg orithm" are frequently used in the reports, and it is claimed that these systems are free from human bias. However, these claims are not supported by detailed explanations of how the algorithms work. Meta's use of quantified transparency (45.37) demonstrates that the platform's transparency strategy is based on numerical data. Metrics such as the number of removed pieces of content, community standards violation rates, and proactive detection rates are highlighted in reports. For example, Meta reports frequently include statements like "562 million pieces of content removed in Q4 2025." This quantification strategy leads to transparency being equated with numerical data. Lectio Socialis 17 Meta's behavioral data extraction discourse (5.89) reflects that the platform's business model is based on user data. Terms such as "user data," "personal data," "advertising," and "profiling" are used in reports, but detailed explanations of how this data is collected and used are not provided. Meta acknowledges the use of user data for advertising purposes but fails to ensure transparency in this process. Islam (2025) examines Meta’s algorithmic transparency and competitive advantage strategy in relation to European data privacy and digital services regulations, analyzing how the platform balances regulatory compliance with its commercial interests. Meta ad opts a strategy of protecting its trade secrets while meeting the transparency obligations imposed by the DSA. This strategy aims to protect its competitive advantage by avoiding disclosing the details of its algorithmic systems. These findings relate prim arily to Algorithmic Neutrality Claim and Quantified Transparency categories and directly address RQ1. Meta emphasizes AI systems and quantitative moderation metrics as indicators of transparency but provides only limited information about how algorithmic decision - making processes operate. This reflects one of the main findings of the study: transparency is primarily communicated through measurable reporting practices rather than through explanations of how algorithmic decisions are made. X: Emphasis on quantification and automated enforcement The X platform has the highest frequency in the quantified transparency category (79.47). X's transparency reports heavily focus on numerical data, highlighting metrics such as the number of removed content items, suspended accounts, and violation rates. T erms like "total," "count," "rate," and "%" are frequently used in the reports. X's quantification strategy equates transparency with measurable metrics. X's automated enforcement discourse (17.72) indicates that the platform's moderation strategy is based on automation. Terms such as "automated," "automatically," "suspended," and "removed" are frequently used in reports. X emphasizes that content moderatio n is largely carried out by automated systems but does not provide detailed explanations of how these systems work. X's claim to algorithmic neutrality (6.28) is lower than other platforms. X emphasizes AI systems but does not assert their neutrality as strongly as Meta. X's discourse on transparency focuses more on procedural compliance and quantified data. X's discourse on political subjectivity (15.96) indicates that the platform legitimizes content moderation through policies and rules. Terms such as "policy," "rules," and "guidelines" are frequently used in reports. X emphasizes that content moderation de cisions are based on community rules, but does not explain how these rules are determined or changed. The findings for X highlight the combined role of Quantified Transparency and Automated Enforcement in addressing RQ1. X emphasizes quantitative moderation metrics and automated enforcement but provides only limited information about how automated enforcem ent decisions are made. This finding suggests that transparency is communicated primarily through measurable reporting practices rather than how automated decisions are made. YouTube: Political subjectivity and automated enforcement YouTube has the highest frequency in the quantified transparency category (92.99). YouTube's transparency reports are excessively focused on numerical data, highlighting Torun 18 metrics such as the number of videos removed, violation rates, and proactive detection rates. YouTube's quantification strategy leads to transparency being limited solely to numerical data. YouTube's discourse on political subjectivity (45.44) indicates that the platform adopts a rule - based approach to content moderation. Terms such as "guidelines," "policy," and "policies" are frequently used in reports. YouTube legitimizes its moderation de cisions by emphasizing community guidelines and content policies. YouTube's use of automated enforcement rhetoric (23.78) indicates that its moderation strategy relies on automation. Terms like "automated," "removed," and "terminated" are frequently used in reports. YouTube emphasizes that content moderation is largely c arried out by automated systems, but does not provide detailed explanations of how these systems work. Quintas - Froufe et al. (2024), while analyzing TikTok's policies to protect young audiences against disinformation, note that similar findings apply to YouTube as well. The platforms claim to combat disinformation by highlighting their corporate policies, b ut fail to provide sufficient evidence regarding the effectiveness of these policies. In this case, transparency appears to be closely linked to reporting procedures and policy implementation. YouTube's findings indicate that Quantified Transparency, Political Subjectivity, and Automated Enforcement jointly shape its transparency discourse. Although the reports emphasize community guidelines and policy implementation, they provide only limited i nformation about how moderation decisions are made. This finding suggests that transparency is communicated primarily through reporting practices rather than through explanations of the moderation process. TikTok: Political subjectivity and government request compliance TikTok has the highest frequency in the political subjectivity category (47.13). TikTok's transparency reports heavily emphasize its community guidelines and content policies. Terms such as "guidelines," "policy," and "terms of service" are frequently used in the reports. TikTok emphasizes that its content moderation decisions are based on community standards, but it does not explain how these standards are determined or changed. TikTok's discourse on compliance with government requests (8.88) reflects the platform's response to regulatory pressures. Reports use terms such as "government request," "law enforcement request," and "government removal request." TikTok demonstrates regu latory compliance by highlighting its compliance rates with government requests. TikTok's quantified use of transparency (49.18) indicates that the platform's transparency strategy is based on numerical data. However, TikTok's reports are shorter and less detailed compared to other platforms. TikTok provides basic metrics but does not offer in - depth analysis and explanations. TikTok's transparency discourse is shaped primarily by Political Subjectivity, Government Request Compliance, and Quantified Transparency, directly addressing RQ1. The reports emphasize community policies, government requests, and quantitative reporting bu t provide only limited information about how moderation decisions are made. This pattern suggests that transparency is communicated primarily through regulatory compliance and reporting practices rather than through explanations of the moderation process. Lectio Socialis 19 Comparison of EU DSA and TR 7253: Regulatory asymmetry The DSA and TR 7253 regulatory texts reflect different priorities regarding transparency and accountability. The DSA text stands out in the category of regulatory asymmetry with a frequency of 10.31. The DSA meticulously regulates obligations, risk assessm ents, independent audits, and transparency reports for very large online platforms (VLOPs). The DSA's use of terms such as "very large online platform," "systemic risk," "audit," and "obligation" indicates that the regulatory framework aims to ensure accou ntability in platform governance. The text of TR 7253 shows a frequency of 6.30 in the category of regulatory asymmetry. TR 7253 imposes obligations on social network providers to appoint local representatives, respond to content removal requests, and provide regular reports. However, the scope of TR 7253 is more limited than that of DSA. TR 7253 does not address issues such as the transparency of algorithmic systems, risk assessments, and independent audits. The fact that the DSA shows a frequency of 15.10 in the algorithmic neutrality claim category indicates that the regulatory text emphasizes the transparency and accountability of algorithmic systems. The DSA aims to regulate the algorithmic systems of plat forms by using terms such as “algorithm”, “algorithmic”, “AI”, and “recommender system”. The TR 7253 text, however, shows a frequency of only 0.70 in the algorithmic neutrality category and almost completely excludes algorithmic systems. The regulatory comparison graph given in Figure 6 visualizes the discursive priorities of the DSA and TR 7253 texts. Figure 6. Regulatory comparison: DSA, TR 7253, and platform transparency discourse. Panel A: comparison of DSA and TR 7253. Panel B: all platforms in regulatory - sensitive categories. The comparison of DSA and TR 7253 reveals a significant regulatory asymmetry. DSA shows a frequency of 10.31 and TR 7253 shows 6.30, while social media platforms exhibit a frequency of 0.00 in this category. This finding reveals that regulatory texts addre ss power imbalances in platform governance, but platforms ignore this asymmetry in their transparency reports. DSA exhibits a relatively high value of 15.10 in the algorithmic neutrality claim category. This finding indicates that the DSA text significantly incorporates discourse on the neutrality and accountability of algorithmic systems. TR 7253, on the other han d, shows only a frequency of 0.70 in the algorithmic neutrality claim category, revealing a discursive gap in the national regulatory text regarding algorithmic accountability. Torun 20 TR 7253 exhibits a higher frequency of 16.10 in the quantified transparency category compared to DSA (10.07). This finding indicates that TR 7253 places greater emphasis on quantitative reporting requirements. However, TR 7253 exhibits zero frequency in th e categories of political subjectivity (0.00), automatic sanctions (0.00), behavioral data extraction (0.00), and selective visibility (0.00), revealing that the discursive scope of the national regulatory text is quite limited. Recent studies on the DSA transparency database highlight limitations related to data quality, consistency, and the reliability of self - reported moderation practices (Kaushal et al., 2024; Trujillo et al., 2023; Drolsbach & Pröllochs, 2024). Mishra and Agrawal (2025) examined how the principles of transparency and accountability are operationalized in DSA, revealing the strengths and weaknesses of the regulatory framework. DSA imposes comprehensive transparency obligations on platforms, but ch allenges remain in effectively implementing and monitoring these obligations. The radar graph presented in Figure 7 visualizes the discursive profile differences between pairs of platforms in three separate panels. Figure 7. Radar chart: comparative transparency discourse profiles (per - panel normalized; axis labels show actual values). The first panel of the radar chart (Meta – X comparison) reveals that Meta exhibits a significantly broader profile than X in terms of its claim to algorithmic neutrality. The second panel (YouTube – TikTok comparison) shows that YouTube dominates in terms of quantified transparency and automated enforcement, while TikTok excels in terms of political subjectivity and compliance with state demands. The third panel (DSA – TR 7253 comparison) reveals that DSA displays a multidimensional profile, while TR 7253 is only prominent in terms of quantified transparency and regulatory asymmetry. Discussion The algorithmic transparency paradox The findings related to RQ1 show that platforms construct transparency primarily through quantified reporting and claims of algorithmic neutrality rather than through meaningful explanations of moderation processes. Although platforms present extensive num erical indicators and moderation statistics, they provide only limited information about how algorithmic decisions are made. This creates an algorithmic transparency paradox in which visibility increases through measurable reporting while the decision - maki ng process itself Lectio Socialis 21 remains opaque. In this respect, the findings confirm Ananny and Crawford's (2018) concept of "seeing without knowing," demonstrating that transparency extends the visibility of moderation outcomes without making the underlying decision - making processes in telligible. Transparency should therefore go beyond the disclosure of information and enable meaningful understanding of how algorithmic governance operates. When the findings related to RQ1 are considered as a whole, they demonstrate that transparency rep orts simultaneously fulfill multiple governance functions. These efforts not only document moderation activities but also construct broader narratives regarding accountability, regulatory compliance, and institutional legitimacy. The following discussion situates these findings within the existing literature throu gh a more comprehensive theoretical framework. These findings align with previous research showing that transparency reports provide limited insight into algorithmic processes (Ananny & Crawford, 2018). Papaevangelou and Votta (2025) examine the tension between platform observability and content governance under DSA, highlighting the challenges of balancing scale and nuance. The transparency obligations imposed by DSA necessitate large - scale data sharing by platforms. However, problems exist regarding the quality and interpretability of this data. Large - scale data sharing makes nuanced and contextual analyses difficult and leads to superficial assessments. The Selective visibility category has the lowest frequency among all documents which indicates that platforms systematically conceal their practices of selectively managing content visibility in their transparency reports. When evaluated within the framewo rk of Han's (2015) critique of the transparency society, the platforms' claim of "transparency" is actually the product of a selective visibility strategy: platforms make visible the information that is advantageous to them, while keeping the essence of th eir algorithmic decision - making processes hidden. Han's (2015) critique of the transparency society illuminates this paradox. The imperative of transparency does not dissolve power but reconfigures it: those who control what is made visible exercise a new form of power. Performative compliance and legitimacy construction The findings related to RQ2 show that transparency reports primarily emphasize regulatory compliance rather than substantive accountability. Platforms present compliance with policies, reporting obligations, and regulatory requirements, but provide limited explanation of how moderation decisions are made. This suggests that transparency reports function not only as reporting documents but also as a way of demonstrating compliance with increasing regulatory expectations. As Parsons (2019) points out, volunta rily produced transparency reports are often ineffective. Platforms publish transparency reports partly to respond to regulatory pressures, but these reports do not provide real accountability. The reports offer selective information and highlight data tha t present platforms in a positive light. Surveillance capitalism and behavioral data extraction The findings related to RQ3 show that transparency claims remain selective rather than comprehensive. While platforms provide detailed information about content moderation activities, they disclose much less about behavioral data collection, profiling prac tices, and the commercial use of user data. This finding indicates that transparency is unevenly Torun 22 distributed across different areas of platform governance and that important aspects of algorithmic decision - making remain largely invisible. The analysis of Behavioral Data Extraction discourse reveals a systematic pattern of selective disclosure that reflects the economic logic of surveillance capitalism. Platforms are extensively forthcoming about their content moderation practices which are primarily a cost center, while remaining opaque about their data collection and monetization practices. This asymmetry is not incidental but reflects the fundamental economic interests of platforms in maintaining the opacity of their behavioral data extrac tion operations. Zuboff's (2019) concept of surveillance capitalism provides a critical framework for understanding platforms' transparency discourses. The platforms' actual business model is to generate advertising revenue by collecting users' behavioral data. Transparenc y reports conceal this data extraction process and focus only on visible activities such as content moderation. Han’s (2015) critique of the transparency society explains the paradoxical nature of the platforms’ transparency discourses. Platforms establish an asymmetrical power relationship by claiming to be transparent while monitoring users. Users are unaware of h ow the platforms’ algorithmic systems work, while the platforms know every move of the users. This asymmetry is problematic from the perspective of democratic values. Bucher (2018) analyzes the societal consequences of “if…then” logic by examining the political power and policies of algorithms. Algorithms are presented as seemingly neutral and objective rules, but in reality, they reflect certain values and priorities. The transparency reports of platforms tend to conceal this political dimension by emphasizing the neutrality of algorithms. Napoli (2019) discusses the social responsibilities of platforms by addressing the public interest dimension of social media. Platforms are not only commercial companies but also a vital part of the public communication infrastructure. Therefore, the trans parency and accountability standards of platforms should be higher than those of traditional media organizations. However, current regulatory frameworks do not make platforms sufficiently accountable. The findings corroborate Suzor's (2019) critique of platform opacity: across all three platforms, transparency reports disclose rule enforcement while the processes through which those rules are formulated and revised remain entirely outside the scope of p ublic accountability. Limitations of the DSA Transparency Database The DSA transparency database has taken an important step by requiring platforms to report their content moderation decisions in a standard format. However, the structural limitations of the database reduce its effectiveness. Consistent with the structural limitations documented by Kaushal et al. (2024), the present analysis finds that the DSA database's data quality and consistency problems constrain meaningful comparative assessment of platform transparency. Platforms create massive datasets by tracking a nd recording every user action. This data is analyzed using machine learning algorithms to predict future user behavior. These predicted behaviors are then used to direct targeted advertising and content recommendations. Transparency reports only reveal a small part of this process (content moderation), concealing the actual economic value creation process. Shahi et al. (2025) analyzed one year's worth of data from the DSA transparency database to reveal what the database shows and does not show about platform moderation. Lectio Socialis 23 The authors note that the database provides quantitative data but does not offer sufficient information to evaluate the context, accuracy, and impact of moderation decisions. Qualitative assessments are necessary to complement the quantitative data. Another limitation of this study is that, although the dataset spans the period between 2017 and 2025, the analysis does not differentiate between the periods before and after the implementation of the DSA (2022) and Law No. 7253 (2020). Future research co uld adopt a longitudinal design to examine how transparency reporting practices have evolved following the implementation of these regulatory frameworks. Transparency as a policy axis: The political significance of DSA and Law No. 7253 The high frequencies of DSA and TR 7253 in the regulatory asymmetry category indicate that both regulatory texts address power imbalances in platform governance. However, the discursive priorities of these two texts exhibit significant differences. DSA dev elops a more comprehensive discourse compared to TR 7253 in the categories of algorithmic neutrality claim (15.10), behavioral data extraction (2.42), and user agency restriction (0.34). TR 7253, on the other hand, exhibits almost zero frequency in these categories, revealing that the discursive scope of the national regulatory text is quite limited. Frosio & Geiger’s (2023) characterization of DSA as a tool for asserting digital sovereignty aligns with the findings of this study. The high frequency of DSA in the regulatory asymmetry category reflects, at a discursive level, the European Union's effort to establish political authority over platform companies. Helberger and Samuelson's (2024) interpretation of DSA as the beginning of a global transparency regime, when considered in light of the findings of this study, suggests that DSA represents a signi ficant step forward. Its discursive framework appears to have strong potential for export to other countries. TR 7253's low frequencies in the categories of political subjectivity (0.00), automatic sanctions (0.00), and algorithmic neutrality claim (0.70) reveal a discursive gap in the national regulatory text in critical areas such as algorithmic governance and c ontent moderation. Griffin's (2025) warning about the risk of depoliticization of the systemic risk management approach is particularly relevant in the context of TR 7253's discursive limitations: the national regulation avoids addressing structural issues such as algorithmic accountability and platform governance, remaining limited only to the framework of sanctions and obligations. The comparison between the DSA and Law No. 7253 also reflects differences in their broader regulatory contexts. The DSA was developed within the European Union's broader accountability framework, where platform governance is supported by independent regula tors, civil society organizations, and academic scrutiny. By contrast, Law No. 7253 primarily focuses on content removal procedures and procedural compliance within the national regulatory framework. These differences suggest that the observed discursive p atterns are shaped not only by the wording of the regulations themselves but also by the institutional and regulatory environments in which they operate. Therefore, the differences identified in this study should be interpreted as reflecting two different approaches to platform governance rather than simply two different legal texts. This interpretation is also consistent with Bradford's (2020) concept of the Brussels Effect, which explains how the European Union's regulatory capacity extends beyond its bor ders and encourages platforms to adopt broader transparency and accountability practices. Torun 24 This broader perspective is also consistent with recent discussions emphasizing that platform governance increasingly involves not only regulatory compliance but also the protection of fundamental rights and democratic accountability (Belic et al., 2025). When these findings are considered together, it becomes clear that platform transparency is not merely a process based on presenting information, but also involves a selective structure that determines which information is highlighted and which is relegate d to the background. This suggests that transparency is not a neutral means of explanation, but is directly related to how platforms frame their own activities. Therefore, transparency reports should be evaluated not only based on the data presented, but a lso on the areas that are limitedly included or completely excluded. In this respect, the study points to the importance of addressing the issue of transparency not only in terms of the quantity or level of openness of data, but also in terms of the relati onship between what is visible and what is made invisible. Conclusion This study comparatively analyzes the transparency reports of Meta, X, YouTube, and TikTok platforms within the framework of the European Union Digital Services Act (DSA) and Türkiye's Law No. 7253. Using systematic qualitative content analysis, 65 platform reports and two regulatory documents published between 2017 and 2025 were examined. The results indicate that the platforms' transparency discourses are based on quantif ied data, claims of algorithmic neutrality, and an emphasis on automated enforcement. The results show that platforms rely on quantified transparency while providing limited insight into algorithmic decision - making. Although regulatory frameworks aim to increase accountability, platform reports remain largely procedural. Quantified transparency is the dominant strategy used in the platforms' transparency reports. Discursive strategies differ among the platforms, but all platforms remain limited to demonstrating procedural compliance. Platform reports partially reflect the DSA's priorities but fall short in terms of algorithmic transparency. Platforms' claims of transparency do not guarantee genuine transparency in algorithmic decision - making processes. Transparency reports are limited to a display of procedural compliance, and algorithmic opacity is structural. Platforms claim to be transpare nt by presenting quantified data but fail to explain how the algorithms work, what criteria they use, and how they affect user rights. The theoretical contribution of this study is to demonstrate that platform transparency is a multidimensional phenomenon and that procedural transparency must be complemented by algorithmic transparency and institutional transparency. Leone de Castris’s (2 025) classification of platform transparency types, Han’s (2015) critique of the transparency society, and Zuboff’s (2019) concept of surveillance capitalism offer a powerful framework for understanding the paradoxical nature of platform transparency disco urses. The practical contribution of this study is to show regulators and policymakers the limitations of transparency reports on platforms and to offer recommendations for developing more comprehensive regulatory tools. While the transparency obligations introdu ced by the DSA are an important step, they should be supported by tools such as independent auditing, API access, and data sharing. Türkiye's Law No. 7253 should be strengthened on issues such as transparency and risk assessments of algorithmic systems. Lectio Socialis 25 The results of this study offer several important contributions to the literature on platform transparency. First, the normalized frequency analysis method provides a methodological framework for the systematic analysis of platform transparency discourses by enabling the comparison of texts of varying lengths. Second, the discursive analysis framework, consisting of ten coding categories, allows for a multidimensional assessment of platforms' transparency strategies. Third, the comparative analysis of DSA a nd TR 7253 reveals differences in the discursive priorities of national and supranational regulatory texts. The study has some limitations. Categories based on human judgment in the coding process can be susceptible to interpretive bias; however, the high inter - coder reliability coefficient ( κ = 0.82) mitigates this risk. Furthermore, only the textual content of the platforms' transparency reports was analyzed, excluding visual and multimedia elements. Future research could integrate the quantitative analysis of the DSA transparency database with the qualitative findings of this study to provide a more comprehens ive assessment of platform transparency. Future research could investigate how platforms respond to regulatory pressures by examining how their transparency reports change over time. How users perceive transparency reports and whether these reports provide genuine accountability are also importan t research questions. The effectiveness of independent audits and the improvement of the DSA transparency database should be priority areas for future studies. This study adopted an approach that focuses on the structural and discursive patterns that characterize the discourse of platform transparency as a whole, rather than tracking changes over time. A longitudinal comparison of reporting practices before and a fter DSA (the enactment of the DSA in 2022) constitutes a promising direction for future research. In conclusion, the discourse on transparency on digital platforms is complex and multifaceted. Platforms strategically use transparency rhetoric to meet regulatory requirements, but remain fundamentally opaque in their algorithmic decision - making processes . Simply presenting quantified data is not enough to achieve real transparency. Detailed information must be provided on how algorithmic systems work, which criteria are used, and how they affect user rights. Only in this way can digital platforms serve de mocratic values and gain the trust of users. This study suggests that the transparency practices of digital platforms should be evaluated not only in terms of how much information is presented, but also in terms of how this information is structured and within what limits it is presented. In this fra mework, increasing transparency alone should not be seen as a sufficient goal; instead, the extent to which the presented information contributes to users' understanding of decision - making processes should be questioned. This approach necessitates rethinki ng the transparency discussion through qualitative depth rather than quantitative increase. Future research may examine changes in transparency reporting before and after major regulatory developments and may also consider the role of visual reporting elements. Such studies could contribute to a broader understanding of platform transparency prac tices. Bibliography Ananny, M., & Crawford, K. (2018). Seeing without knowing: Limitations of the transparency ideal and its application to algorithmic accountability. New Media & Society , 20 (3), 973 – 989. https://doi.org/10.1177/1461444816676645 Torun 26 Belic, J., Canfield, M., Griffin, R., Lahmann, H., & Sander, B. (2025). Human rights in the governance of digital platforms: Introduction to the special issue. Transnational Legal Theory , 16 (4), 507 – 518. https://doi.org/10.1080/20414005.2025.2600781 Bradford, A. (2020). The Brussels effect: How the European Union rules the world . Oxford University Press. BTK. (2020). Law No. 7253 on the regulation on social network provider . Republic of Turkiye Ministry of Transport and Infrastructure, Information Technologies and Communications Authority. Bucher, T. (2018). If…Then: Algorithmic power and politics . Oxford University Press. Cobbe, J., Veale, M., & Singh, J. (2021). Reviewable automated decision - making: A framework for accountable algorithmic systems. In Proceedings of the 2021 ACM Conference on Fairness, Accountability, and Transparency (pp. 598 – 609). ACM. https://doi.org/10.1145/3442188.3445921 Dikba ş Torun, E. (2026). Coding scheme and frequency dataset for “Transparency discourse on digital platforms: A comparative textual analysis of platform reports and regulatory texts in the EU and Türkiye” [Data set]. Zenodo. https://doi.org/10.5281/zenodo.21374555 Di Porto, F., & Zuppetta, M. (2021). Co - regulating algorithmic disclosure for digital platforms. Policy and Society , 40 (2), 272 – 293. https://doi.org/10.1080/14494035.2020.1809052 Drolsbach, C., & Pröllochs, N. (2024). Content moderation on social media in the EU: Insights from the DSA Transparency Database. In Proceedings of the ACM Web Conference 2024 (pp. 1 – 10). ACM. https://doi.org/10.1145/3589335.3651482 Elkin - Koren, N., & Perel, M. (2019). Contesting algorithms: Restoring the public interest in content filtering by artificial intelligence. Big Data & Society , 7 (2), 1 – 13. https://doi.org/10.1177/2053951720932296 European Union. (2022). Regulation (EU) 2022/2065 of the European Parliament and of the Council of 19 October 2022 on a Single Market for Digital Services and amending Directive 2000/31/EC (Digital Services Act). Official Journal of the European Union , L 277/1. https://eur - lex.europa.eu/legal - content/EN/TXT/?uri=CELEX:32022R2065 Flyverbom, M. (2019). The digital prism: Transparency and managed visibilities in a datafied world . Cambridge University Press. Foucault, M. (1977). Discipline and punish: The birth of the prison . Pantheon Books. Frosio, G., & Geiger, C. (2023). Taking fundamental rights seriously in the Digital Services Act's platform liability regime. European Law Journal , 29 (1 – 2), 31 – 52. https://doi.org/10.1111/eulj.12475 Fung, A., Graham, M., & Weil, D. (2007). Full disclosure: The perils and promise of transparency . Cambridge University Press. Gillespie, T. (2018). Custodians of the internet: Platforms, content moderation, and the hidden decisions that shape social media . Yale University Press. Google LLC. (2019 – 2025). YouTube community guidelines enforcement report . Google Transparency Report. https://transparencyreport.google.com Gorwa, R. (2019). What is platform governance? Information, Communication & Society , 22 (6), 854 – 871. https://doi.org/10.1080/1369118X.2019.1573914 Griffin, R. (2025). Governing platforms through corporate risk management: The politics of systemic risk in the Digital Services Act. European Law Open , 4 (2), 223 – 253. https://doi.org/10.1017/elo.2025.17 Lectio Socialis 27 Grimmer, J., & Stewart, B. M. (2013). Text as data: The promise and pitfalls of automatic content analysis methods for political texts. Political Analysis , 21 (3), 267 – 297. https://doi.org/10.1093/pan/mps028 Han, B. - C. (2015). The transparency society . Stanford University Press. Husovec, M. (2024). The Digital Services Act's red line: What the Commission can and cannot do about disinformation. Journal of Media Law , 16 (1), 47 – 56. https://doi.org/10.1080/17577632.2024.2362483 Ilieva, M. (2023). Tackling online disinformation – EU platform regulations. Yearbook of the Department of Mass Communication , 10 , 145 – 162. https://doi.org/10.33919/ydmc.23.10 Islam, S. (2025). Algorithmic transparency and competitive advantage: Meta’s strategic response to European data privacy and digital services regulation. Journal of Digital Policy & Regulation , 3 (1), 45 – 67. Kaushal, R., van de Kerkhof, J., Goanta, C., Spanakis, G., & Iamnitchi, A. (2024). Automated transparency: A legal and empirical analysis of the Digital Services Act Transparency Database. In Proceedings of the 2024 ACM Conference on Fairness, Accountability, and Transparency (pp. 1 – 15). ACM. https://doi.org/10.1145/3630106.3658960 Kuckartz, U. (2014). Qualitative text analysis: A guide to methods, practice and using software . SAGE. Landis, J. R., & Koch, G. G. (1977). The measurement of observer agreement for categorical data. Biometrics , 33 (1), 159 – 174. https://doi.org/10.2307/2529310 Leerssen, P. (2024). Outside the black box: From algorithmic transparency to platform observability in the Digital Services Act. European Journal of Law and Technology , 14 (2), 1 – 28. Leone de Castris, A. (2025). Types of platform transparency: An analysis of discourse around transparency and global digital platforms. Public Integrity , 27 (3), 340 – 354. https://doi.org/10.1080/10999922.2024.2304741 Lodder, A. R., & Carvalho, J. M. (2022). Online platforms: Towards an information tsunami with new requirements on moderation, ranking and traceability. European Law Review , 47 (3), 412 – 435. https://doi.org/10.54648/eulr2022025 Mayring, P. (2014). Qualitative content analysis: Theoretical foundation, basic procedures and software solution . Klagenfurt. Meta Platforms. (2017 – 2025). Transparency reports . https://transparency.meta.com Mishra, N., & Agrawal, B. (2025). Transparency and accountability in the Digital Services Act: A critical assessment. SSRN Electronic Journal . https://doi.org/10.2139/ssrn.5671992 Mittelstadt, B. (2016). Auditing for transparency in content personalization systems. International Journal of Communication , 10 , 4991 – 5002. Napoli, P. M. (2019). Social media and the public interest: Media regulation in the disinformation age . Columbia University Press. Papaevangelou, C., & Votta, F. (2025). Trading nuance for scale? Platform observability and content governance under the DSA. Internet Policy Review , 14 (3), 1 – 24. https://doi.org/10.14763/2025.3.2037 Parsons, A. (2019). The ineffectiveness of voluntarily produced transparency reports. Business & Society , 58 (1), 103 – 131. https://doi.org/10.1177/0007650317717957 Piovano, A. (2025). Automating DSA enforcement: A socio - technical framework for transparency compliance. Zenodo . https://doi.org/10.13135/2785 - 7867/12271 Quintas - Froufe, N., González - Neira, A., & Fiaño - Salinas, C. (2024). Corporate policies to protect against disinformation for young audiences: The case of TikTok. Frontiers in Communication , 9 , Article 1410100. https://doi.org/10.3389/fcomm.2024.1410100 Torun 28 Reid, A., & Ringel, E. (2025). Digital intermediaries and transparency reports as strategic communications. The Information Society , 41 (2), 91 – 109. Roberts, S. T. (2019). Behind the screen: Content moderation in the shadows of social media . Yale University Press. Samuelson, P., & Helberger, N. (2024). The Digital Services Act as a global transparency regime . SSRN Working Paper. https://papers.ssrn.com/sol3/papers.cfm?abstract_id=4783341 Shahi, G. K., Tessa, B., Trujillo, A., & Cresci, S. (2025). A year of the DSA Transparency Database: What it (does not) reveal about platform moderation during the 2024 European Parliament election. arXiv preprint arXiv:2504.06976 . https://doi.org/10.48550/arXiv.2504.06976 Strowel, A. (2024). The Digital Services Act: Transparency as an efficient tool to curb the spread of disinformation on online platforms. Zenodo . https://doi.org/10.5281/zenodo.10618280 Suzor, N. (2019). Lawless: The secret rules that govern our digital lives . Cambridge University Press. Terzis, P. A., Veale, M., & Gaumann, N. (2024). Law and the emerging political economy of algorithmic audits. In Proceedings of the 2024 ACM Conference on Fairness, Accountability, and Transparency (pp. 1 – 12). ACM. https://doi.org/10.1145/3630106.3658970 TikTok. (2020 – 2025). Transparency reports . https://www.tiktok.com/safety/transparency Trujillo, A., Fagni, T., & Cresci, S. (2023). The DSA Transparency Database: Auditing self - reported moderation actions by social media. arXiv preprint arXiv:2312.10269 . https://doi.org/10.48550/arXiv.2312.10269 Urman, A., & Makhortykh, M. (2023). How transparent are transparency reports? Comparative analysis of transparency reporting across online platforms. Telecommunications Policy , 47 (3), 102477. Wei, J. T. - Z., Zufall, F., & Jia, R. (2024). Operationalizing content moderation “accuracy” in the Digital Services Act. In Proceedings of the 2024 AAAI/ACM Conference on AI, Ethics, and Society (pp. 1 – 10). ACM. https://doi.org/10.1609/aies.v7i1.31744 X Corp. (2018 – 2024). Global transparency report . https://transparency.x.com Zuboff, S. (2019). The age of surveillance capitalism: The fight for a human future at the new frontier of power . PublicAffairs. Notes on contributors Emel Dikba ş Torun is an Associate Professor and the founding head of the Department of New Media and Communication at Pamukkale University, Faculty of Communication. She received her PhD from Hacettepe University and has conducted academic research at Purdue University in t he United States. Her research focuses on new media, digitalization, and communication studies. ORCID Emel Dikba ş Torun https://orcid.org/0000 - 0002 - 7882 - 9295 Author contributions The author confirms sole responsibility for all aspects of this study. Emel Dikba ş Torun conceptualized the research, developed the theoretical framework, conducted the analysis, Lectio Socialis 29 and wrote and revised the manuscript. AI - assisted tools were used solely for language editing and proofreading and did not contribute to the study design, analysis, or interpretation of the findings. Conflict of interest statement The author declares no conflicts of interest. Funding The author received no financial support for the research, authorship, and/or publication of this article. Data availability statement The primary data analyzed in this study consist of 65 publicly available transparency reports published by Meta, X, YouTube, and TikTok (2017 – 2025), as well as the texts of the EU Digital Services Act (DSA, EU Regulation 2022/2065) and Türkiye’s Law No. 72 53; these are publicly available third - party documents, and their sources are cited in the reference list. The coding scheme (comprising the category definitions and keyword lists that form the basis of Table 1) and the normalized frequency dataset underly ing Figures 3 – 7 are made publicly available under a CC BY 4.0 licence within the Lectio Socialis Zenodo community at https://doi.org/10.5281/zenodo.21374555 . Ethics approval statement This study analyzes publicly available secondary data — transparency reports issued by Meta, X, YouTube, and TikTok, and the texts of the European Union Digital Services Act and Turkish Law No. 7253. It does not involve human participants, animal subjects, or the collection of primary data from individuals. Accordingly, ethics committee approval was not required for this study.