Art. 40 DSA and research data access in the German federal election: a mixed experience at best
Article 40 of the Digital Services Act (DSA) aims at enabling research data access and strengthening transparency in the context of very large online platforms. However, the reality of our research in the run-up of the German federal election were stalled applications, highly restrictive and limited data access, insufficient documentation, unresponsive support and technical problems with service reliability.

With the Digital Services Act (DSA), the European Union has taken a decisive step toward increasing the transparency of large online platforms. For researchers examining systemic risks associated with digital platforms—such as any actual or foreseeable negative effects on civic discourse and electoral processes, and public security—this represents a turning point. Article 40 of the DSA establishes a legal framework that grants vetted researchers access to data from very large online platforms (VLOPs) and very large online search engines (VLOSEs) to study their impact on the public sphere.
The German federal election in February 2025 marked an important test of Article 40’s implementation. Access to research data has never been more urgent. Today, elections are contested as much on social media feeds and algorithmic timelines as in mass media and traditional campaign rallies. Yet, despite the ambitious promises of the DSA, obtaining research data remains an uphill battle—hindered by bureaucratic hurdles, restrictive access conditions, and platform reluctance. This report outlines our experiences as the Social Media Observatory, a project within the Research Institute Social Cohesion and the Leibniz-Institute for Media Research, in studying platform dynamics regarding issues like polarization, disinformation, and extremism during a national election. While utilizing other data sources—such as scraping and open APIs for traditional online media and platforms like Telegram—our project aimed to rely on Article 40 of the DSA to access META platforms (Facebook and Instagram) as well as TikTok.
Meta Data Access
Gaining access to Meta’s data proved to be a frustrating experience. Meta has partnered with the Inter-university Consortium for Political and Social Research (ICPSR) at the University of Michigan to facilitate researcher access to public data from Facebook and Instagram through the Social Media Archive (SOMAR). In theory, this collaboration should allow qualified researchers to utilize the Meta Content Library and its API within SOMAR’s secure Virtual Data Enclave. It serves as the successor to CrowdTangle, the now-defunct service previously widely used by researchers and journalists to access Meta content.
In practice, however, SOMAR paused all applications for data access from December 2024 to March 2025—an unfortunate timing, given that the German federal election took place on February 23, 2025. When we reached out to SOMAR, citing the DSA and inquiring about alternative access options and potential timelines, we were merely referred to the ongoing suspension of applications. No alternative access options were offered, and we were further informed that the application process could take up to an additional eight weeks.
Given that this timeline made it impossible for us to conduct our research close to the election—when timely access was most critical due to the sensitivity of our research topics—we reached out to the German national coordinator. This administrative position, established under the DSA, is responsible for overseeing compliance, facilitating researcher-platform cooperation, and ensuring transparent data access.
We do not know exactly what triggered this, but SOMAR reopened our inquiry in mid-February and referred us to a team of Meta transparency officers, who then met with us to discuss alternative ways to access platform data. While they apologized for the difficulties we had faced and appeared genuinely committed to finding a solution, this engagement only occurred shortly before the election. As a result, conducting research during the critical pre-election period was effectively impossible.
Even worse, the provided Meta Content Library offered only highly restricted data access. For example, it allowed access to Facebook pages through a web interface and provided downloadable data only for pages with more than 15,000 followers or a verified badge. Since the majority of German parliament members—both at the federal level and Bundesländer levels—do not meet this threshold and many lack a verified badge, meaningful quantitative research based on this data was effectively impossible.
Additionally, our access will be temporary, requiring us to reapply (and once again invest time and resources) to obtain permanent access. This is especially frustrating as just half a year ago, the shutdown of Crowdtangle—the primary tool for accessing Meta platform data— rendered years of work by developers and researchers unusable and irreproducible.
TikTok Data Access
TikTok data access presented a slightly different, but no less frustrating, experience. While our application process proceeded relatively smoothly, working with the API proved to be highly challenging. The TikTok Research API has been available since 2023 (Sato, 2023) and has already faced criticism for its data quality (Darius, 2024).
The main issues we encountered while using the API included restrictive rate limits, insufficient documentation, unresponsive support for navigating these limitations, and problems with service reliability.
Restrictive rate limits: The API is highly restrictive, limiting the total number of records that can be obtained to 10,000 per day and prohibiting further collation or merging of the data. Actual video data is inaccessible through the API, and the research API’s Terms of Service (TOS) explicitly prohibit the use of the data for purposes like scraping video content.
These restrictions are deeply problematic, as they effectively prevent meaningful quantitative research into the actual content of TikTok communication, undermining key research goals, particularly in sensitive areas such as disinformation and extremism studies.
The rate limits are even more severe for research focusing on comments below TikTok videos. As mentioned, each API endpoint returns only 10,000 records per day—a number easily exceeded by the volume of comments on videos from relevant politicians. This limitation severely hampers research into phenomena such as inauthentic behavior.
Moreover, finding and implementing workarounds to bypass these restrictions is both complicated and time-consuming, and often violates the TOS).
Insufficient documentation: In addition to the challenges outlined, the documentation for the research API is sparse and imprecise, especially regarding the API’s limitations. For instance, the API includes a query endpoint that allows users to search for content matching specific criteria, such as posts from certain accounts or containing particular keywords. However, the documentation fails to specify how many criteria can be included in a single query or what the consequences are for exceeding this limit.
Inconsistent API behavior: Given our research objectives of tracking a wide range of public entities leading up to the German federal election, the limited number of queries allowed per day forced us to formulate larger queries to request queried to request data for multiple actors. In practice, the API responded to these queries with error codes and inconsistent behavior. While we could successfully request data for approximately 100 accounts in a single query, the API returns paged results for 100 videos per reply, with the option to scroll through the entire result set—if only the scrolling function worked consistently.
Non-responsive support: In summary, we encountered unpredictable and inconsistent behavior from the API and detailed the issue to TikTok’s support contacts via email and the ticket system at the end of November 2024. Unfortunately, we received no response, leaving us without a resolution to these critical issues.
Service reliability: In addition to the API’s limitations and documentation issues, we encountered problems with service reliability. During our work with the API, the service occasionally failed. Such interruptions are problematic, as they compromise the quality of the data collection. For instance, incidents of manipulation efforts or hate speech that occur during these breaks—especially those deleted by moderation or by the originators themselves before data collection resumes—may be lost and cannot be recovered.
Finally, the default tier of research access does not include access to the data clean room, which is essential for conducting thorough and secure research on large-scale sensitive data not accessible via the API.
Research access in the context of Art. 40 DSA: A mixed experience at best
Overall, our experiences highlight that while the DSA is a significant and important step forward in ensuring data access for independent systemic risk research, its implementation so far has fallen short of its promises. The non-responsiveness of the platforms, along with restrictive and complicated access—such as the challenges we faced with TikTok, which unfortunately mirror those encountered by other researchers who previously had access to the META Content Library—often create the impression of obstruction by design – which they hopefully are not.
As researchers, our time and resources are mostly publicly funded, making them precious and limited. The scale and urgency of the research needed to address the challenges our democracies face as outlined above could not be greater. It is high time the platforms fulfill their responsibility and enable data access the DSA mandates, so researchers such as can focus their attention on researching and understanding these digital risks to our societies instead of maneuvering bureaucratic hurdles and technical difficulties.
Authors
Jan Rau is a research associate in the Media Research Methods Lab at the Leibniz-Institute for Media Research | Hans-Bredow-Institut, working on its (Social) Media Observatory, a project for the Research Institute Social Cohesion. He is an advisory member of the Center for Communication, Media and Information (ZeMKI) at the University of Bremen and is pursuing his Ph.D. at the Lab for Digital Communication and Information Diversity (DCID) under Prof. Dr. Cornelius Puschmann. His main interest lies in online-extremism, polarization and platform governance.
Philipp Kessling is a research associate at the Hamburg sub-institute of the RISC. He is an advisory member of the Center for Communication, Media and Information (ZeMKI) at the University of Bremen and is pursuing his Ph.D. at the Lab for Digital Communication and Information Diversity (DCID) under Prof. Dr. Cornelius Puschmann. His main interest lies in digital public spheres and the interplay of platforms.
Gregor Wiedemann is a Senior Researcher for Computational Social Science at the Hamburg RISC location. His research focuses on, among other things, the automation of discourse and content analysis using text mining and natural language processing, as well as the automatic detection of hate speech in social media.
Felix Victor Münch is an interdisciplinary researcher in computational social science. He currently works as a Postdoc Researcher at the Leibniz-Institute for Media Research | Hans-Bredow-Institut in Hamburg, where he is a Principal Investigator of the (Social) Media Observatory (SMO), which he co-founded, co-conceptualized, and co-built as a research facilitator and service infrastructure for the Research Institute Social Cohesion (RISC). With a B.Sc. in Physics (LMU, Munich, Germany), an M.A. in Journalism (LMU and German Journalist School, Munich, Germany), and work experience in online media brand communication, his main fields of interest are network science, online (social) media, natural language processing, and theories regarding the networked public sphere.
References
Darius, P. (2024, September 24). Researcher Data Access Under the DSA: Lessons from TikTok’s API Issues During the 2024 European Elections | TechPolicy.Press. Tech Policy Press. https://techpolicy.press/-researcher-data-access-under-the-dsa-lessons-from-tiktoks-api-issues-during-the-2024-european-elections
Sato, M. (2023, February 21). Researchers will get access to TikTok data—Pending company approval. The Verge. https://www.theverge.com/2023/2/21/23604737/tiktok-research-api-expansion-public-user-data-transparency
OpenEdition schlägt Ihnen vor, diesen Beitrag wie folgt zu zitieren:
Jan Rau, Philipp Kessling, Gregor Wiedemann, Felix Victor Münch (21. Mai 2025). Art. 40 DSA and research data access in the German federal election: a mixed experience at best. Zusammenhalt begreifen. Abgerufen am 5. März 2026 von https://doi.org/10.58079/13ze0
