I. Introduction
Transparency has long been regarded as a cornerstone of accountability and trust in both democratic governance and corporate regulation. Rooted in the right to know, it empowers citizens and consumers to scrutinise decision-making and ensure compliance when self-regulation fails.Footnote 1 In the digital age, transparency has become indispensable to platform accountability, as content moderation profoundly shapes online discourse yet remains opaque.Footnote 2 Public, academic, and regulatory calls for transparency have remained constant in platform governance but their lack of specificity has also enabled platforms to frame selective disclosures as compliance tools rather than meaningful oversight mechanisms.Footnote 3 To become a vector for accountability, transparency must be meaningful: receivers of transparency must be able to interpret, understand, and act upon the information disclosed.Footnote 4 Yet, even when transparency mechanisms are in place, their effectiveness depends on the reliability of the information they disclose. In the context of platform governance, the central challenge is not transparency per se, but verifiable, meaningful transparency that allows independent confirmation of disclosed practices rather than mere observation.
Existing disclosure practices largely enable platforms to describe their content moderation systems while leaving external actors without tools to assess their accuracy. The transparency requirements in the Digital Services Act (DSA) were designed to address precisely this issue. Conceived as a “transparency machine,” the DSA sought to unveil the opaque practices of content moderation that had long avoided public scrutiny and to position itself as a cornerstone in the emerging architecture of platform governance.Footnote 5
Yet, the first transparency reports fell short of these expectations. The absence of a standardised reporting framework produced disclosures that were inconsistent, difficult to interpret, and ultimately incapable of realising the DSA’s ambition of generating meaningful transparency.Footnote 6 More importantly, the framework lacked mechanisms to verify the accuracy and completeness of the information provided.Footnote 7 Indeed, the DSA’s auditing obligations are limited to Very Large Online Platforms and Search Engines (VLOPs/VLOSEs) and focus primarily on procedural adequacy rather than the substantive accuracy of reported data.Footnote 8 For all other intermediaries, verification rests with oversight authorities empowered to request additional information or initiate investigations but the practical feasibility of exercising such powers across hundreds of intermediary services is doubtful.Footnote 9
Resource constraints, jurisdictional fragmentation, and competing enforcement priorities make comprehensive verification of platforms’ disclosures virtually unattainable. This structural limitation effectively preserves platforms’ discretion to curate their self-reporting, thereby undermining the reliability of the transparency regime.Footnote 10 Paradoxically, the DSA already provides the technical infrastructure necessary for verification. The Database of Statements of Reasons contains decision-level data that, if linked to the aggregate figures disclosed by platforms under Articles 15 and 24, could enable systematic consistency checks and independent validation.Footnote 11 The 2025 harmonisation reformsFootnote 12 to the transparency reporting templates and the Database schema could have facilitated such alignment, shifting transparency from a self-reporting exercise to a verifiable system of accountability. Whether these reforms will succeed in converting transparency from descriptive to evidentiary is the central focus of this article.
Part II will discuss how the initial framework for transparency requirements under the DSA perpetuated the reliability problem in content moderation transparency reporting, whereby platforms provide an illusion of transparency by regularly publishing information about their practices, but avoid accountability by preventing independent verification of their disclosures. Part III analyses the Implementing Regulation and the harmonised Database of Statements of Reasons schema to assess whether the reforms have addressed the reliability problem identified in Part II. Part IV discusses the underutilisation of the regulatory tools in the DSA transparency schema and argues that it stems not from its technical incapacity but from the absence of regulatory design for verification, ultimately identifying the legal, technical and institutional preconditions required for the Database to evolve into a verification tool capable of reinforcing the reliability of the DSA transparency framework as a whole. Part V concludes.
II. The initial DSA framework and its reliability problem
Before assessing the reforms introduced by the 2025 Implementing RegulationFootnote 13 , it is necessary to examine the original structure of the DSA’s transparency regime and the reasons why it failed to produce verifiable information. The DSA’s first implementation cycle (2023–2024) dramatically expanded the amount of data that intermediary services were required to disclose but left unresolved the central question of reliability. This section argues that the DSA’s architecture combined extensive reporting duties with narrowly circumscribed verification mechanisms, resulting in a transparency system that generated visibility without accountability.
1. Mandatory transparency obligations under the DSA
The DSA introduced a graduated framework of transparency obligations that scales in intensity with a provider’s size, function, and societal impact.Footnote 14 Articles 15, 24 and 42 set out the core reporting requirements, while Articles 37 and 40 create limited avenues for verification through audits and researcher access.
At a minimum, all providers of intermediary services must publish annual transparency reports detailing orders from Member States’ authorities, notices submitted under the Article 16 notice-and-action mechanism, own-initiative content moderation actions, complaints received through internal complaint-handling systems, and the use of automated tools. For all these categories, platforms must include details on type of allegedly infringing content, action taken, median time needed to react, as well as use of automated systems.
Online platforms face additional obligations under Article 24, including the disclosure of user-base metrics and outcomes of out-of-court dispute settlements.Footnote 15 VLOPs and VLOSEs must report every six months and include further detail on their moderation workforce, including linguistic expertise, training protocols, and indicators of accuracy.Footnote 16
Although this architecture increased the quantity and periodicity of data available, it offered few means to verify accuracy and completeness. Only VLOPs and VLOSEs are subject to annual independent audits under Article 37, where auditors must assess procedural compliance. For all other intermediaries, oversight remains confined to the general investigatory powers of the Commission and national Digital Services Coordinators. In practice, these authorities lack the capacity to review hundreds of reports across the EU. The result is an asymmetric system in which the majority of intermediaries publish unverified data while the most powerful actors undergo audits focused primarily on processes rather than outcomes.Footnote 17
While Articles 15 to 24 determine what platforms must report, Article 17 establishes how they must document individual moderation decisions. The interaction – or lack thereof – between these two instruments will prove decisive for the reliability of the overall framework.
2. The DSA database and transparency-reporting limitations
Among the DSA’s transparency instruments, the Database of Statements of Reasons (hereafter, Database) represents its most innovative and potentially transformative component. Since February 2024, all online intermediaries have been required to log in this centralised repository every “statement of reasons” (SoR) issued to users whose content or accounts has been affected by a content moderation decision. The Database provides, for the first time, insights into individual decisions by recording, for each action, the relevant facts, the legal or contractual basis, the use of automated means during the detection and decision-making processes and the restriction applied.
The Database is governed by articles 17, 24 and 43. Article 17 requires that platforms provide users with a clear and specific statement about the content moderation decision and possibilities for redress. Article 24(5) obliges providers to submit each statement “without undue delay” to the Commission’s Database in machine-readable form, and Article 43 establishes the funding system for the Database maintenance. The Database’s scope, however, is not exhaustive and excludes deceptive high-volume commercial content, government-ordered removals and law-enforcement-sensitive actions.
Technically, the Database operates through an application programming interface (API) that allows single or bulk submissions and enforces predefined attributes for content type, restriction category, detection method, territorial scope and legal basis, standardising documentation and enabling automated data feeding at scale.Footnote 18 Nevertheless, an analysis of the initial design of the Database and the SoR template revealed limitations that constrained its potential as accountability tool.Footnote 19 The original Database’s taxonomy diverged from platforms’ classifications in transparency reports, precluding cross-dataset comparison; redress information required under Article 17(3)(f) was not captured; and entries could not be amended to reflect reversals or appeals. Above all, the 2024 schema included no legal or technical mechanism to link decision-level data with the aggregate figures reported under Articles 15 and 24.
As a result, the initial Database functioned primarily as a descriptive repository rather than a verification instrument. While it greatly expanded visibility into moderation activity, it did not allow regulators or researchers to test the accuracy of platforms’ aggregate disclosures. Only with the 2025 harmonisation process did the Commission begin addressing these structural deficiencies – a reform analysed in Section III. The following section delves into how this early disconnect between datasets exemplified and perpetuated the reliability problem inherent in the DSA’s first transparency reporting cycle.
3. The reliability problem
The reliability problem arises when transparency practices allow platforms to shape the narrative concerning their content moderation processes while preventing external validation of the claims they advance.Footnote 20 The selective structure of transparency reports, where companies determine what to disclose and how to frame it, has long hindered the efficacy of voluntary transparency practices, transforming disclosures into instruments of reputational management that project the appearance of openness while remaining substantively evasive or incomplete.Footnote 21 Where legal obligations are absent or weakly enforced, platforms face no material consequences for omissions or inaccuracies, visibility is rewarded and disclosure is treated as an end in itself.Footnote 22
Even after the DSA introduced mandatory reporting, the reliability problem persisted. The DSA sanctions mainly the failure to report, not the quality of reported information, thereby allowing formal compliance to coexist with substantive opacity.Footnote 23 In the absence of any systematic verification processes to assess the veracity and completeness of platforms’ transparency disclosures, platforms retain discretion to curate what they reveal, omitting or downplaying information that could invite criticism.Footnote 24
Auditing obligations and researcher access to internal data – the few verification elements in the DSA transparency framework – are limited to VLOPs and VLOSEs, and, even there, focused primarily on procedural adequacy rather than factual correctness. For the majority of online platforms, oversight is dependent on authorities constrained by limited resources and competing enforcement priorities. Inaccurate or strategically curated reports may therefore go undetected unless they attract targeted regulatory scrutiny. The result is a transparency regime that expands the quantity of publicly available information while leaving its reliability uncertain and an uncaptured potential. As the first analyses of the DSA Database demonstrate, selective or inconsistent disclosures remain common even under a binding legal framework.Footnote 25
Unless transparency mechanisms are coupled with verifiability, disclosure risks reinforcing platforms’ narrative control rather than constraining it. It is within this context that the Database of Statements of Reasons assumes particular importance: properly leveraged, it could supply the empirical foundation for cross-checking aggregate reports and thereby transform transparency from a descriptive obligation into an evidentiary one. The next section examines whether the 2025 reforms – through the Implementing Regulation and the harmonised Database schema – have addressed the reliability problem.
III. The Implementing Regulation: reforms and expectations
Adopted on 4 November 2024, the Implementing Regulation laying down templates for transparency reporting (Implementing Regulation) represents the EU’s first systematic attempt to standardise how intermediary services disclose information under Articles 15, 24 and 42 DSA.Footnote 26 The Commission presented the reform as a “step towards harmonised and comparable reports,” seeking to replace the fragmented and heterogeneous disclosures that characterised the DSA’s initial implementation cycle.Footnote 27
By introducing common reporting deadlines, an exhaustive taxonomy of categories and uniform data-formatting rules, the Implementing Regulation aims to limit platforms’ discretion and render disclosures comparable both across services and over time.Footnote 28 These changes also resulted in parallel updates to the template for SoR submissions to the Database, theoretically aligning the granular data stored in the Database with the aggregated figures disclosed in transparency reports. Whether this harmonisation can also render them verifiable, however, remains an open question.
1. Key changes to transparency reporting requirements
The Implementing Regulation introduces a comprehensive set of harmonisation measures designed to transform the DSA’s transparency reporting framework from a loose collection of self-defined disclosures into a standardised, machine-readable system. The first reform standardises the temporal scope of transparency reports.Footnote 29 This alignment resolves one of the most practical barriers to cross-platform comparability. During the first three reporting cycles, platforms selected divergent reference periods, producing temporal gaps that rendered cross-platform comparison particularly challenging.Footnote 30 Reports describing overlapping but non-identical timeframes captured moderation activity from inconsistent intervals, frustrating attempts to establish baselines or identify sector-wide trends.Footnote 31 The harmonised reporting periods now ensure that disclosures refer to the same temporal windows, allowing regulators, civil-society organisations and researchers to conduct horizontal comparisons across services and longitudinal assessments over time. Temporal harmonisation thus eliminates a structural ambiguity in the DSA’s initial framework, though it remains a procedural, rather than substantive, improvement.
The second reform concerns the standardisation of the substantive categories used to classify moderation actions. In the previous iterations of transparency reports, platforms’ divergent categorisation approaches prevented meaningful cross-platform comparison and verification.Footnote 32 The Implementing Regulation now mandates the use of a common and exhaustive list of high-level categories across both transparency reports and the Database. When reporting on actions based on alleged illegality with EU or national law, platforms are required to use the category of illegality indicated by competent authority.Footnote 33 Similarly, for notices reported under the notice-and-action mechanism, platforms must rely on the category corresponding to the explanation provided by the notifier.Footnote 34
When reporting on own initiative content moderation action, platforms must use the high-level categories of illegality or the violation of the terms and conditions established in an exhaustive list.Footnote 35 Additionally, for illegal content categories, platforms have the opportunity to provide contextual information about their interpretation of the categories on the basis of their internal categorisation practices, specifying how each illegal content category description relates to the platforms’ own terms and conditions policies. Platforms can also create new sub-categories to provide further context, but the Commission clarified that new sub-categories cannot be “used as a catch-all categories.”Footnote 36
By aligning the taxonomy used in transparency reports with the categories applied in the Database, the Implementing Regulation establishes for the first time a consistent vocabulary across aggregate and decision-level disclosures.Footnote 37 This structural alignment theoretically lays the groundwork for cross-mechanism verification, enabling regulators to test whether the aggregate figures reported under Articles 15 and 24 correspond to the individual moderation decisions logged in the Database.
Beyond harmonising reference periods and substantive categories, the Implementing Regulation standardises the format of transparency reports, prescribing not only what platforms must disclose but how this information must be presented.Footnote 38 Additionally, the Implementing Regulation also defines uniform data-typing conventions, limiting the potential for variations in units and denominations that made the first DSA reports virtually incomparable.Footnote 39
Collectively, the prescribed templates eliminate the interpretive discretion that previously allowed platforms to comply formally with Article 15 DSA while presenting incomparable data. These changes bring long-needed procedural coherence to the DSA’s transparency regime, ensuring that platforms report within the same timeframes, use the same categorical vocabulary, and present information in the same technical format. Yet, while transparency reporting templates and harmonised rules enhanced consistency and comparability, platforms’ disclosures remain self-generated, perpetuating the reliability problem.
2. Database template modifications
The second pillar of the 2025 reform concerns the Database of Statements of Reasons structure, which the Commission updated to ensure alignment with the new transparency-reporting templates entering into force on 1 July 2025. The revision of the SoR template and of the API schema governing submissions align the data categories used in the Database with those applied in transparency reports, creating a continuous reporting chain from individual decisions to aggregate statistics. In principle, this harmonisation could transform the Database from a stand-alone repository into a component of an integrated transparency framework.Footnote 40
The SoR template follows (at most) the requirements under Article 17, translating its qualitative requirements into mostly pre-defined attributes and machine-readable fields.Footnote 41 The 2025 revision reorganises these attributes into a clearer hierarchy covering (i) information about the affected content, (ii) the type, duration and territorial scope of the restriction, (iii) the factual basis and origin of the investigation, (iv) the use of automated means, (v) the legal or contractual ground, and (vi) the categorisation of the violation.Footnote 42 Three aspects of this reorganisation are particularly consequential for assessing whether the Database can contribute to verification rather than achieve pure documentation.
First, the section on decisional grounds now requires platforms to specify, in all cases, whether a moderation action is based on illegality or incompatibility with the provider’s terms and conditions. When a decision relies on a finding of illegality, platforms must indicate the precise legal provision invoked (illegal_content_legal_ground) and provide an accompanying explanation of its applicability (illegal_content_explanation).Footnote 43 When the restriction stems from a breach of contractual terms, they must cite the relevant section of the terms of service (incompatible_content_ground) and justify why the content violates it (incompatible_content_explanation).Footnote 44 These are now mandatory fields, replacing what had previously been optional or inconsistently filled. The new guidance clarifies that platforms must refer to the exact law or contractual clause relied upon and explain its relevance to the decision.Footnote 45 This formulation represents a meaningful step forward, as it formally prevents generic statements such as “incompatible with T&Cs” and compels reference to the specific provision invoked. Nevertheless, the reform continues to depend on self-declared accuracy. Even with more prescriptive fields, providers may comply nominally while maintaining opacity through minimal or vague legal citations (“Article 10 law,” “Section 3 policy violation”) that defy meaningful verification.Footnote 46 Furthermore, fields that could force further specificity – such as the reference URL allowing platforms to link directly to the cited legal or contractual provision – remain optional (decision_ground_reference_url).Footnote 47 The first analyses of the Database already highlighted the risk in leaving this field optional, with research showing that platforms rarely provide precise references for non-mandatory fields.Footnote 48
Second, the template retains the structure governing the disclosure of the facts and circumstances relied upon in taking a moderation decision. The free-text field (decision_facts) remains mandatory, complemented by the structured indicator identifying the source of the investigation (source_type), i.e., whether the decision originated from a user notice under Article 16, a trusted flagger, another external notification, or the platform’s own initiative.Footnote 49 Because the revision leaves the substance of the requirement unchanged, the limitations observed in the Database’s first iteration may persist. Analyses of early entries found that, although the “decision_facts” field was mandatory and central to contextualising content moderation decisions, it was mostly filled with generic or templated responses that offered little meaningful insight.Footnote 50
Third, the categorisation of violations has undergone the most visible restructuring. The Database now mirrors the taxonomy established in Annex II to the Implementing Regulation, requiring each statement of reasons to include one mandatory high-level category (statement_category) and, optionally, additional categories (statement_category_addition) and specifications (category_specification).Footnote 51 This exhaustive, harmonised list spans seventeen high-level categories and over eighty subcategories. Providers may still create new subcategories under “keyword_other,” but must supply a textual description of the content type and cannot add new high-level categories.Footnote 52 This system is intended to prevent the proliferation of idiosyncratic classifications that, as previously documented, hindered cross-platform comparability and frustrated the objective of “meaningful transparency” under Article 15 DSA.Footnote 53 However, this controlled flexibility does not prevent the risk for abuse. Over-reliance on open subcategories could reintroduce fragmentation by allowing providers to label edge cases under self-defined headings (“other misinformation,” “borderline nudity”) that escape harmonised scrutiny.Footnote 54
Despite these improvements, several omissions identified in earlier analyses of the Database structure remain unresolved. The submission template still lacks fields corresponding to the redress mechanisms available to affected users, even though Article 17(3)(f) explicitly requires that statements of reasons include “clear and user-friendly information on the possibilities for redress available.” Without this element, however, the Database captures only the moment of restriction, rather than the broader procedural relationship between the platform and the user. It therefore provides no means to assess whether users are informed of their right to contest a decision, how platforms operationalise internal complaint mechanisms, or whether redress avenues are offered consistently across jurisdictions and content types. Likewise, the Database still provides no field to record whether a decision was later challenged through an internal complaint or out-of-court settlement, nor any mechanism to update or withdraw an SoR once a decision has been reversed.Footnote 55 Consequently, what is recorded in the Database is a static snapshot of initial decisions rather than the full lifecycle of a moderation decision, providing an incomplete and potentially misleading picture of platform practices.
Overall, the 2025 revisions improve the internal coherence of the Database template and strengthen its alignment with the transparency reporting framework. Yet these gains remain incremental rather than transformative. The Database continues to serve primarily as a documentation tool rather than a verification mechanism. The reforms ensure that platforms describe their decisions in a uniform manner, but they do not ensure that those descriptions are complete or accurate. In other words, harmonisation mitigates inconsistency but not unreliability. The Database now speaks the same language as transparency reports, yet it still cannot independently confirm whether what platforms say is true.
IV. The database’s untapped potential as a verification tool
The 2025 Implementing Regulation and the accompanying updates to the Statement of Reasons (SoR) template represent a substantial advancement toward a more harmonised transparency regime. Despite these improvements, the DSA’s transparency framework remains essentially descriptive. The Database continues to operate as a repository of disclosures designed for observation, not verification, leaving the underlying reliability problem unsolved. As in the first generation of DSA transparency reports, accuracy remains presumed rather than demonstrated. This limitation is fundamental. The Database now records billions of individual content moderation decisions, but these granular records remain disconnected from the aggregated data presented in transparency reports. Without a formalised mechanism to reconcile the two, regulators cannot determine whether the figures disclosed in transparency reports accurately reflect the respective moderation activity.
This article argues that, if properly leveraged, the Database could provide the systematic verification necessary for meaningful transparency, by allowing regulators to cross-check reported aggregated figures against individual SoR submissions, detect inconsistencies, and prioritise investigative resources. This would not only enhance accountability for VLOPs and VLOSEs, already subject to audits, but – perhaps more importantly – extend verifiability to the broader ecosystem of intermediaries that fall outside the auditing framework but must still comply with the same transparency obligations.
This section explains how the Database’s current underutilisation stems not from its technical incapacity but from the absence of regulatory design for verification. It proceeds in three steps. Section IV.1 examines how the Database’s architecture and the recent template updates have improved its harmonisation but left structural disconnections unaddressed. Section IV.2 discusses why the verification function is particularly crucial for smaller intermediaries and Digital Services Coordinators (DSCs). Section IV.3 identifies the legal, technical, and institutional preconditions required for the Database to evolve into a verification tool capable of reinforcing the reliability of the DSA transparency framework as a whole.
1. From documentation to verification
The Implementing Regulation marks a turning point in the DSA’s transparency framework. For the first time, transparency reports and statements of reasons share compatible, machine-readable structures. Yet a significant gap remains: there is no mechanism linking individual SoRs to the aggregated data in transparency reports, nor any means of automatically testing the consistency between the two datasets.
The Implementing Regulation nevertheless brings the system within reach of verifiability. First, the insistence on machine-readability creates a data architecture that can support automated cross-analysis. Second, Annex II’s exhaustive taxonomy of illegal-content categories (seventeen high-level types, each subdivided into specific sub-categories) aligns the categorisation logic of transparency reports with that of the Database. Building on this comparability structure, a limited set of additional obligations would suffice to transform the Database from a passive documentation tool into an active oversight mechanism. These obligations would focus on introducing traceability across the two transparency mechanisms: (i) mandatory cross-referencing fields linking aggregated figures to Database submissions; (ii) update fields in SoRs for corrected or appealed decisions; and (iii) automatic aggregation checks at the time of report submission. Absent such requirements, however, the Database will remain a purely descriptive exercise rather than a verifiable transparency system.
2. Why verification matters beyond VLOPs/VLOSEs
The DSA’s auditing obligations apply only to VLOPs and VLOSEs. These entities are subject to independent annual audits under Article 37, which evaluate their compliance with risk-assessment and mitigation duties, including the accuracy of their transparency reporting. By contrast, all other intermediaries remain outside the auditing framework. They are nonetheless bound by the same Article 15 transparency obligations and therefore continue to self-report moderation data without external verification.
Integrating verification functions within the Database would help fill this gap. Because all intermediaries are required to submit SoRs, the Database already collects granular data capable of supporting basic reliability checks. Automated consistency analyses – performed either by the Commission or by Digital Services Coordinators (DSCs) – could compare the volume, distribution, and categorisation of individual decisions against the aggregated statistics reported under Articles 15 and 24. Even simple anomaly-detection tools could flag inaccurate patterns, such as missing categories, inaccurate totals, or inconsistent breakdowns across reporting mechanisms.Footnote 56
Embedding automated consistency checks within the Database infrastructure could reveal discrepancies at the time of submission, allowing DSCs to flag inconsistencies and request further information. Such checks would not replicate the depth of Article 37 audits but would create a proportional verification layer for non-VLOPs. They would enable regulators to prioritise oversight where discrepancies are most pronounced and, over time, incentivise systemic accountability across the intermediary ecosystem. Database-based verification would extend the logic of auditability to all providers, transforming formal compliance into meaningful transparency.
3. Structural preconditions for verification
Transforming the Database into a functional verification instrument would require three interdependent preconditions: technical, legal and institutional. For what concerns the technical foundation, the harmonisation process has already created the structural compatibility necessary for automated reconciliation between transparency reports and Database submissions. This harmonisation ensures that both instruments are expressed in compatible, machine-readable structures and that their variables can be compared automatically. If regulators were to analyse these aligned datasets in parallel, they could – theoretically – detect inconsistencies between what platforms report in aggregate and what they record at the decision level. The technical precondition for verification is therefore already in place but without a traceability requirement, these features will remain underutilised in oversight.
On the other hand, the legal precondition for verification is absent from the DSA transparency framework. Neither the DSA nor the Implementing Regulation explicitly require providers to ensure internal consistency between their transparency reports and the SoRs transmitted to the Database. Article 17 imposes a duty to communicate explanations for individual moderation decisions, while Articles 15 and 24 require the publication of aggregate transparency data, but the two obligations operate independently. Similarly, the Implementing Regulation focuses on harmonising format and taxonomy, not on establishing mandatory coherence or cross-verification. As a result, even though the datasets are now structurally comparable, platforms are under no legal obligation to demonstrate that their reported figures accurately reflect the decisions recorded in the Database, and regulators lack a clear legal mandate to test such consistency. Introducing verification would therefore require an interpretative clarification or secondary measure establishing a traceability requirement, i.e., the expectation that aggregated data must be reconcilable with the underlying statements of reasons. Without such an interpretative step, verification remains permissible in theory but not enforceable in practice.
Finally, the verification process would require institutional capacity. The DSA already distributes oversight responsibilities across the Commission and Digital Services Coordinators, empowering both with investigative and information-request powers.Footnote 57 On the other hand, there is no structured process to prioritise their use of those powers. For VLOPs and VLOSEs, independent audits provide an initial verification cycle. On the contrary, for smaller intermediaries, oversight remains reactive, triggered only by complaints or ad hoc suspicion of non-compliance. Establishing an intermediate verification process could make this oversight more systematic. Ideally, when platforms submit their annual transparency reports, the figures disclosed would be automatically matched to the statements of reasons they uploaded to the Database during the same reporting period.Footnote 58 This could occur through an interface integrated with the Database, allowing DSCs to view a high-level consistency summary for each provider. Platforms showing unexplained discrepancies would be flagged for further review.
Such a system would not create new investigatory powers but would help regulators use existing ones more strategically, enabling proactive oversight and evidence-based prioritisation of enforcement efforts.
4. Mandating traceability
The preceding analysis indicates that the harmonisation process already laid the foundation for the verification of accuracy of transparency reports through cross-comparison with Database entries. To operationalise the structural preconditions identified in Section IV.3, a legal basis linking the different transparency instruments of the DSA into a coherent oversight framework is needed. Addressing this gap does not necessarily require a redesign of the Database or the creation of a new reporting layer. Rather, it calls for the introduction of an explicit traceability requirement establishing that the aggregated information disclosed under Articles 15 and 24 must be reconcilable with the statements of reasons transmitted under Article 17.
This requirement could be articulated through a delegated act under Article 15 or Article 24, both of which empower the Commission to specify the form, content and details of reporting obligations. Such an instrument could clarify that traceability forms part of the duty of accuracy inherent in transparency reporting, thereby allowing verification to occur within the existing legal framework.
A traceability requirement would not expand the substantive scope of the DSA’s reporting duties. Providers would not be required to submit additional information but to ensure that the data they already transmit remain internally coherent across mechanisms. The obligation could be framed procedurally: when publishing transparency reports, providers must be able, upon request by DSCs or the Commission, to demonstrate how the reported figures align with the SoRs submitted during the relevant reporting period. This clarification would give legal effect to the semantic alignment introduced by the Implementing Regulation, transforming it from a technical feature into a compliance benchmark. It would also provide a defined legal basis for regulators to perform consistency analyses and to request methodological explanations where anomalies arise.
Formally articulating the relationship between Articles 15, 17 and 24 would also support a more coherent use of existing investigatory powers. Under Articles 51–53, both the Commission and DSCs may already request information and conduct inspections, but these powers operate reactively. A traceability requirement would create an intermediary verification stage, allowing DSCs to use the Database as a first-line screening tool before initiating formal proceedings. Platforms whose reported data diverge significantly from their Database entries could be prioritised for investigation or asked to provide clarifications under Article 52(1). This would strengthen proportionality and resource allocation in enforcement while maintaining the DSA’s graduated oversight structure.
V. Conclusion
The July 2025 reforms mark a significant step toward a coherent transparency framework under the DSA. By harmonising temporal scopes, taxonomies and reporting formats, the Implementing Regulation and the corresponding updates to the Database template address many of the inconsistencies that undermined the first transparency cycle. These measures enhance comparability and coherence across instruments, creating the structural preconditions for a verifiable transparency regime. Yet, the reliability problem – the persistent inability to verify whether the information platforms disclose accurately reflects their moderation practices – remains the central weakness of the DSA’s transparency regime. The DSA already contains the tools and structure necessary to address this issue: the Database of Statements of Reasons provides granular, decision-level data that could enable systematic verification of reported figures. However, the 2025 reforms stop short of operationalising this potential. The Database remains a powerful but underutilised instrument, technically capable of supporting verification but limited to an observation tool.
Closing this gap would not require a new reporting layer or a fundamental redesign of the existing mechanisms. A narrowly tailored duty of traceability could ensure that the data reported under Articles 15 and 24 are reconcilable with the statements of reasons transmitted under Article 17. Such a clarification would translate the harmonisation achieved through the 2025 reforms into operational verifiability, allowing regulators to assess whether the reported aggregates correspond to the underlying decision-level data. Traceability would not expand the scope of reporting but would give legal effect to the expectation of internal coherence already implicit in the DSA’s transparency obligations.
Embedding traceability would enhance both proportionality and oversight. For VLOPs and VLOSEs, it would complement existing auditing duties by facilitating independent cross-checks; for smaller intermediaries, it would introduce a baseline verification mechanism within the same legal framework. In this way, traceability would extend accountability beyond the audited entities, turning transparency into a reliable measure of platform conduct rather than a self-declared account of it. The DSA’s transparency architecture would thus evolve from one that records platform behaviour to one that can substantively verify it, realising its original purpose as an instrument of trustworthy and enforceable accountability.
Open access
