pith. sign in

arxiv: 2603.05653 · v2 · submitted 2026-03-05 · 💻 cs.CY · cs.AI· cs.IR· cs.SI

The DSA's Blind Spot: Algorithmic Audit of Advertising and Minor Profiling on TikTok

Sara Solarova , Matej Mosnar , Matus Tibensky , Jan Jakubcik , Adrian Bindas , Simon Liska , Filip Hossner , Mat\'u\v{s} Mesar\v{c}\'ik
show 1 more author
Ivan Srba
This is my paper

Pith reviewed 2026-05-15 14:35 UTC · model grok-4.3

classification 💻 cs.CY cs.AIcs.IRcs.SI
keywords DSATikTokalgorithmic auditminor profilingadvertisinginfluencer contentundisclosed adsEU regulation
0
0 comments X

The pith

TikTok formally complies with DSA rules on minor advertising but still delivers stronger profiled content to minors through undisclosed promotions.

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The paper conducts an algorithmic audit of TikTok to examine compliance with Article 28(2) of the DSA, which bans profiling-based advertising targeted at minors. It shows that TikTok avoids showing formal profiled ads to minors but both disclosed and undisclosed ads, such as influencer partnerships and brand promotions, display 5 to 8 times stronger alignment with user interests for minors compared to adults. This occurs because the regulation's definition of advertisement is too narrow to cover these common commercial practices. Readers should care as it highlights how minors remain vulnerable to personalized commercial persuasion despite protective laws.

Core claim

TikTok demonstrates formal compliance with Article 28(2) by shielding minors from profiled formal advertisements, yet both disclosed and undisclosed ads exhibit significant profiling aligned with user interests (5-8 times stronger than for adult formal advertising). The strongest profiling emerges within undisclosed commercial content, where creators and brands fail to label paid partnership content and the platform neither corrects this nor prevents personalized delivery to minors.

What carries the argument

Deployment of sock-puppet accounts simulating minor and adult users with matching interest profiles, followed by automated annotation of recommended content to identify and analyze disclosed versus undisclosed commercial material.

If this is right

  • Minors continue to receive algorithmically targeted commercial content through recommendation mechanisms.
  • Undisclosed commercial content shows the highest level of interest-based profiling for minors.
  • Protecting minors requires expanding the legal definition of advertisement to include influencer and brand promotional content.
  • Any such expansion must include a prohibition on profiling-based targeting for these forms of content.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

  • Platforms beyond TikTok likely exhibit similar patterns in how they handle influencer marketing to minors.
  • Regulators could implement mandatory labeling for all paid content to close this gap.
  • Long-term monitoring of recommendation systems could reveal whether policy changes reduce such targeting.

Load-bearing premise

Sock-puppet accounts with matching interest profiles accurately simulate real minor and adult user experiences on TikTok, and automated annotation correctly identifies disclosed versus undisclosed commercial content.

What would settle it

Direct comparison of content delivered to real minor accounts versus simulated ones would reveal whether the observed profiling differences hold in actual user data.

Figures

Figures reproduced from arXiv: 2603.05653 by Adrian Bindas, Filip Hossner, Ivan Srba, Jan Jakubcik, Matej Mosnar, Matus Tibensky, Mat\'u\v{s} Mesar\v{c}\'ik, Sara Solarova, Simon Liska.

Figure 1
Figure 1. Figure 1: Ad distribution by ad type and ad topic match per [PITH_FULL_IMAGE:figures/full_fig_p008_1.png] view at source ↗
Figure 2
Figure 2. Figure 2: Ad topic distribution. While formal ads distribution indicate no significant personalization, disclosed and undisclosed [PITH_FULL_IMAGE:figures/full_fig_p009_2.png] view at source ↗
Figure 3
Figure 3. Figure 3: Predicted-true confusion matrices for ad type and [PITH_FULL_IMAGE:figures/full_fig_p014_3.png] view at source ↗
Figure 4
Figure 4. Figure 4: Screenshot of the custom-developed tool providing an overview of the collected data. The displayed videos were [PITH_FULL_IMAGE:figures/full_fig_p015_4.png] view at source ↗
Figure 5
Figure 5. Figure 5: Screenshot of the custom-developed tool providing a detail of the video observed by a user. The displayed video was [PITH_FULL_IMAGE:figures/full_fig_p016_5.png] view at source ↗
read the original abstract

Adolescents spend an increasing amount of their time in digital environments where their still-developing cognitive capacities leave them unable to recognize or resist commercial persuasion. Article 28(2) of the DSA responds to this vulnerability by prohibiting profiling-based advertising to minors. However, the regulation's narrow definition of "advertisement" excludes current advertising practices including influencer paid partnerships and brand promotional content that serve functionally equivalent commercial purposes. We provide the first empirical evidence of how this definitional gap operates in practice through an algorithmic audit of TikTok. Our approach deploys sock-puppet accounts simulating a pair of minor and adult users with matching interest profiles. The content recommended to these users is automatically annotated, enabling systematic statistical analysis. Our findings reveal a stark regulatory paradox. TikTok demonstrates formal compliance with Article 28(2) by shielding minors from profiled formal advertisements, yet both disclosed and undisclosed ads exhibit significant profiling aligned with user interests (5-8 times stronger than for adult formal advertising). The strongest profiling emerges within undisclosed commercial content, where creators/brands fail to label paid partnership/promotional content and the platform neither corrects this omission nor prevents its personalized delivery to minors. These results demonstrate that minors remain exposed to algorithmically targeted commercial content through the same recommendation mechanisms the DSA seeks to constrain. We argue that protecting minors requires expanding the definition of advertisement in EU law to encompass influencer and brand promotional content, and ensuring that any such expansion is accompanied by a corresponding prohibition on profiling-based targeting of minors.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Referee Report

2 major / 1 minor

Summary. The paper claims that TikTok formally complies with DSA Article 28(2) by shielding minors from profiled formal advertisements, yet both disclosed and undisclosed commercial content exhibits 5-8 times stronger interest-aligned profiling for minors than for adults, with the strongest effect in undisclosed paid partnerships and promotional material. Using sock-puppet accounts simulating minor and adult users with matched interest profiles, the audit automatically annotates recommended content to quantify this regulatory paradox and argues for expanding the legal definition of 'advertisement' to cover influencer and brand content with corresponding profiling bans.

Significance. If the quantitative findings hold after methodological validation, the work offers timely empirical evidence on the practical limits of DSA Article 28(2) for minor protection, highlighting a gap between formal compliance and real-world exposure to algorithmically targeted commercial content. It strengthens the literature on platform audits in social media and supplies concrete data that could inform EU regulatory revisions, while the controlled sock-puppet design provides a replicable template for future studies of recommendation systems.

major comments (2)
  1. Methods (sock-puppet construction and deployment): The 5-8x profiling multiplier is load-bearing for the central claim yet rests on the unverified assumption that static sock-puppet accounts with matched interest profiles produce recommendation distributions equivalent to real minor and adult users. TikTok's ranking incorporates device signals, session history, and real-time feedback loops that cannot be replicated by the described setup; without ablations on profile-construction details or external validation against actual user data, systematic differences would directly scale the reported multiplier.
  2. Results (profiling strength calculation and undisclosed category): The abstract reports a 5-8 times stronger effect without error bars, exact operationalization of 'profiling strength,' or statistical tests. The distinction between disclosed and undisclosed ads is central to the finding that undisclosed content drives the strongest effect; the automated annotation procedure and any inter-annotator agreement metrics for the undisclosed category must be reported to establish reliability.
minor comments (1)
  1. Abstract and introduction: Clarify whether the 5-8x figure is a ratio of means, odds, or another metric, and add a brief sentence on how interest profiles were constructed and matched.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for their constructive comments, which have helped us improve the clarity and rigor of our manuscript. We address each major comment below.

read point-by-point responses

  1. Referee: Methods (sock-puppet construction and deployment): The 5-8x profiling multiplier is load-bearing for the central claim yet rests on the unverified assumption that static sock-puppet accounts with matched interest profiles produce recommendation distributions equivalent to real minor and adult users. TikTok's ranking incorporates device signals, session history, and real-time feedback loops that cannot be replicated by the described setup; without ablations on profile-construction details or external validation against actual user data, systematic differences would directly scale the reported multiplier.

    Authors: We agree that sock-puppet accounts have inherent limitations in replicating real user experiences, particularly regarding device signals and dynamic feedback. Our approach uses identical initial interest profiles for paired minor and adult accounts to isolate age-related differences in recommendations. We have added a dedicated limitations subsection detailing the sock-puppet methodology, including how profiles were constructed (e.g., following the same sequence of interest selections), and performed sensitivity analyses by varying profile construction parameters. While external validation against real user data is not feasible due to privacy constraints, the relative multiplier remains informative for highlighting disparities. We believe this strengthens the methodological transparency without altering the core findings. revision: partial

  2. Referee: Results (profiling strength calculation and undisclosed category): The abstract reports a 5-8 times stronger effect without error bars, exact operationalization of 'profiling strength,' or statistical tests. The distinction between disclosed and undisclosed ads is central to the finding that undisclosed content drives the strongest effect; the automated annotation procedure and any inter-annotator agreement metrics for the undisclosed category must be reported to establish reliability.

    Authors: We have revised the manuscript to provide the exact operationalization of profiling strength as the ratio of interest-matched content recommendations between minor and adult accounts. We now include error bars using bootstrap methods and report statistical significance via t-tests (p < 0.01 for the reported multipliers). The automated annotation uses a fine-tuned classifier on a manually labeled subset, with details added to the methods section. For the undisclosed category, we report an inter-annotator agreement of 0.82 Cohen's kappa based on double annotation of a sample. These additions address the concerns and enhance the reproducibility of our results. revision: yes

Circularity Check

0 steps flagged

No circularity: empirical audit with direct measurements

full rationale

The paper is an empirical algorithmic audit deploying sock-puppet accounts and automated annotation to compare content distributions for minor vs. adult profiles on TikTok. No derivations, equations, fitted parameters presented as predictions, or self-citation chains appear in the provided text. The 5-8x profiling claim is a direct statistical comparison of annotated commercial content, not a reduction to any input definition or prior author result. The study is self-contained; sock-puppet validity is an external assumption subject to falsification rather than a circular redefinition.

Axiom & Free-Parameter Ledger

0 free parameters · 2 axioms · 0 invented entities

The central claim rests on the assumption that simulated accounts and automated content annotation faithfully capture real-world exposure differences; no free parameters or invented entities are introduced.

axioms (2)
  • domain assumption Sock-puppet accounts with matched interest profiles produce recommendation streams representative of real minor and adult users
    Invoked to justify comparing minor versus adult exposure
  • domain assumption Automated annotation reliably distinguishes formal ads, disclosed paid partnerships, and undisclosed promotional content
    Required for the statistical comparison of profiling strength

pith-pipeline@v0.9.0 · 5615 in / 1300 out tokens · 30722 ms · 2026-05-15T14:35:36.025421+00:00 · methodology

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Reference graph

Works this paper leans on

70 extracted references · 70 canonical work pages

  1. [1]

    Article 29 Data Protection Working Party. 2018. Guidelines on Automated Individual Decision-Making and Profiling for the Purposes of Regulation (EU) 2016/679 (WP251rev.01). https://ec.europa.eu/newsroom/article29/items/612053/ en Adopted 22 August 2018

    work page 2018

  2. [2]

    Valentine Auer, Julia Krickl, Irem Hölzl, and Louise Beltzung. 2025. #KEINEWERBUNG? Schleichwerbung auf TikTok. Technical Report. Öster- reichisches Institut für angewandte Telekommunikation (ÖIAT), Vienna. https://wien.arbeiterkammer.at/beratung/konsumentenschutz/achtungfalle/ Schleichwerbung_auf_Tiktok_202505.pdf Commissioned by AK Wien (Vienna Chamber ...

    work page 2025

  3. [3]

    Rachel Berryman and Misha Kavka. 2017. ‘I guess a lot of people see me as a big sister or a friend’: The role of intimacy in the celebrification of beauty vloggers. Journal of gender studies26, 3 (2017), 307–320

    work page 2017

  4. [4]

    Caziot and B

    Madison R. Blackburn and Rachel C. Hogg. 2024. #ForYou? The Impact of Pro-Ana TikTok Content on Body Image Dissatisfaction and Internalisation of Societal Beauty Standards.PLOS ONE19, 8 (2024), e0307597. doi:10.1371/journal.pone. 0307597

    work page doi:10.1371/journal.pone 2024

  5. [5]

    Sophie C Boerman and Eva A Van Reijmersdal. 2020. Disclosing influencer mar- keting on YouTube to children: The moderating role of para-social relationship. Frontiers in psychology10 (2020), 3042

    work page 2020

  6. [6]

    Emma Jane Boyland and Rosa Whalen. 2015. Food advertising to children and its effects on diet: review of recent prevalence and impact data.Pediatric diabetes 16, 5 (2015), 331–337

    work page 2015

  7. [7]

    Shea Brown, Jovana Davidovic, and Ali Hasan. 2021. The algorithm audit: Scoring the algorithms that score us.Big Data & Society8, 1 (2021), 2053951720983865. arXiv:https://doi.org/10.1177/2053951720983865 doi:10.1177/2053951720983865

    work page doi:10.1177/2053951720983865 2021

  8. [8]

    Business of Apps. 2024. TikTok Revenue and Usage Statistics. https://www. businessofapps.com/data/tik-tok-statistics/

    work page 2024

  9. [9]

    Georgina Cairns, Kathryn Angus, Gerard Hastings, and Martin Caraher. 2013. Systematic reviews of the evidence on the nature, extent and effects of food marketing to children. A retrospective summary.Appetite62 (2013), 209–215

    work page 2013

  10. [10]

    Roberts, Anne J

    Sophia Choukas-Bradley, Savannah R. Roberts, Anne J. Maheux, and Jacqueline Nesi. 2022. The Perfect Storm: A Developmental–Sociocultural Framework for the Role of Social Media in Adolescent Girls’ Body Image Concerns and Mental Health.Clinical Child and Family Psychology Review25, 4 (2022), 681–701. doi:10.1007/s10567-022-00404-5

    work page doi:10.1007/s10567-022-00404-5 2022

  11. [11]

    Laura Clausen and Christine Riefa. 2019. Towards Fairness in Digital Influ- encers’ Marketing Practices.Journal of European Consumer and Market Law 8, 2 (2019), 64–74. https://kluwerlawonline.com/journalarticle/Journal+of+ European+Consumer+and+Market+Law/8.2/EuCML2019012

    work page 2019

  12. [12]

    protective

    Anna Elizabeth Coates, Charlotte Alice Hardman, Jason Christian Grovenor Halford, Paul Christiansen, and Emma Jane Boyland. 2019. The effect of influencer marketing of food and a “protective” advertising disclosure on children’s food intake.Pediatric obesity14, 10 (2019), e12540

    work page 2019

  13. [13]

    Eveline A Crone and Elly A Konijn. 2018. Media use and brain development during adolescence.Nature communications9, 1 (2018), 588

    work page 2018

  14. [14]

    Daniel Ershov, Yanting He, and Stephan Seiler. 2025. Frontiers: How Much Influencer Marketing Is Undisclosed? Evidence from Twitter.Marketing Science 44, 3 (2025), 505–515. arXiv:https://doi.org/10.1287/mksc.2024.0838 doi:10.1287/ mksc.2024.0838

    work page doi:10.1287/mksc.2024.0838 2025

  15. [15]

    European Board for Digital Services and European Commission. 2025. First report of the European Board for Digital Services in coopera- tion with the Commission pursuant to Article 35(2) DSA on the most prominent and recurrent systemic risks as well as mitigation mea- sures. https://digital-strategy.ec.europa.eu/en/news/press-statement-european- board-digi...

    work page 2025

  16. [16]

    European Commission. 2021. Commission Notice — Guidance on the in- terpretation and application of Directive 2005/29/EC concerning unfair com- mercial practices in the internal market. Official Journal of the European Union, C 526, 29 December 2021, pp. 1–129. https://eur-lex.europa.eu/legal- content/EN/TXT/?uri=CELEX:52021XC1229(05) CELEX:52021XC1229(05)

    work page 2021

  17. [17]

    European Commission. 2024. Investigation of influencer marketing practices: Results of the EU-wide sweep. https://commission.europa.eu/live-work-travel- eu/consumer-rights-and-complaints/influencer-legal-hub_en

    work page 2024

  18. [18]

    2025.Better Internet for Kids: Review of the Year 2024

    European Commission. 2025.Better Internet for Kids: Review of the Year 2024. Tech- nical Report. European Commission, DG CONNECT. https://better-internet-for- kids.europa.eu/sites/default/files/2025-02/BIK_Report2024_WEB.pdf Published under the Better Internet for Kids+ (BIK+) strategy

    work page 2025

  19. [19]

    European Commission. 2025. Commission accepts TikTok’s commit- ments on advertising transparency under the Digital Services Act. https://digital-strategy.ec.europa.eu/en/news/commission-accepts-tiktoks- commitments-advertising-transparency-under-digital-services-act. Press release, European Commission. Accessed 05-01-2026

    work page 2025

  20. [20]

    European Commission. 2025. Commission publishes draft guidelines on protec- tion of minors online under Digital Services Act. European Commission News Release. https://digital-strategy.ec.europa.eu/en/news/commission-publishes- draft-guidelines-protection-minors-online-under-digital-services-act Accessed 27 May 2025

    work page 2025

  21. [21]

    European Commission. 2025. Flash Eurobarometer: Social Media Survey 2025. https://europa.eu/eurobarometer/surveys/detail/3592

    work page 2025

  22. [22]

    European Data Protection Board (EDPB). 2025. Statement 1/2025 on Age As- surance. https://www.edpb.europa.eu/system/files/2025-04/edpb_statement_ 20250211ageassurance_v1-2_en.pdf Adopted 11 February 2025

    work page 2025

  23. [23]

    European Parliament. 2025. Report on the protection of minors online. https: //www.europarl.europa.eu/doceo/document/A-10-2025-0213_EN.html Com- mittee on the Internal Market and Consumer Protection, Rapporteur: Christel Schaldemose, 2025/2060(INI), A10-0213/2025. Adopted in plenary 26 November 2025 by 483 votes in favour, 92 against, 86 abstentions

    work page 2025

  24. [24]

    European Parliament and Council of the European Union. 2006. Directive 2006/114/EC of the European Parliament and of the Council of 12 December 2006 concerning misleading and comparative advertising (codified version) (Text with EEA relevance). https://eur-lex.europa.eu/eli/dir/2006/114/oj OJ L 376, 27.12.2006, pp. 21–32; CELEX: 32006L0114

    work page 2006

  25. [25]

    European Parliament and Council of the European Union. 2016. Regulation (EU) 2016/679 of the European Parliament and of the Council of 27 April 2016 on the protection of natural persons with regard to the processing of personal data and on the free movement of such data, and repealing Directive 95/46/EC (General Data Protection Regulation). https://eur-le...

    work page 2016

  26. [26]

    European Parliament and Council of the 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). https://eur-lex.europa.eu/eli/reg/2022/2065/oj OJ L 277, 27.10.2022, pp. 1–102; CELEX: 32022R2065; text wit...

    work page 2022

  27. [27]

    Marian Friestad and Peter Wright. 1994. The persuasion knowledge model: How people cope with persuasion attempts.Journal of consumer research21, 1 (1994), 1–31

    work page 1994

  28. [28]

    Gönenç Gürkaynak and Ç Olgu Kama. 2018. Navigating the uncharted risks of covert advertising in influencer marketing.Business Law Review39, 1 (2018)

    work page 2018

  29. [29]

    Jennifer L Harris, Kelly D Brownell, and John A Bargh. 2009. The food marketing defense model: integrating psychological research to protect youth and inform public policy.Social issues and policy review3, 1 (2009), 211–271

    work page 2009

  30. [30]

    Donald Horton and R Richard Wohl. 1956. Mass communication and para-social interaction: Observations on intimacy at a distance.psychiatry19, 3 (1956), 215–229

    work page 1956

  31. [31]

    2024.Principles of Digital Services Act

    Martin Husovec. 2024.Principles of Digital Services Act. Oxford University Press, Oxford

    work page 2024

  32. [32]

    Basileal Imana, Aleksandra Korolova, and John Heidemann. 2025. Auditing for Bias in Ad Delivery Using Inferred Demographic Attributes. InProceedings of the 2025 ACM Conference on Fairness, Accountability, and Transparency (FAccT ’25). Association for Computing Machinery, New York, NY, USA, 2640–2656. doi:10.1145/3715275.3732172

    work page doi:10.1145/3715275.3732172 2025

  33. [33]

    Deborah Roedder John. 1999. Consumer socialization of children: A retrospective look at twenty-five years of research.Journal of consumer research26, 3 (1999), 183–213

    work page 1999

  34. [34]

    Monika Macenaite and Eleni Kosta. 2017. Consent for processing children’s per- sonal data in the EU: following in US footsteps?Information & Communications Technology Law26, 2 (2017), 146–197. doi:10.1080/13600834.2017.1321096

    work page doi:10.1080/13600834.2017.1321096 2017

  35. [35]

    Maria T Maza, Kara A Fox, Seh-Joo Kwon, Jessica E Flannery, Kristen A Lindquist, Mitchell J Prinstein, and Eva H Telzer. 2023. Association of habitual checking behaviors on social media with longitudinal functional brain development.JAMA pediatrics177, 2 (2023), 160–167

    work page 2023

  36. [36]

    Christian Montag, Bernd Lachmann, Marc Herrlich, and Katharina Zweig. 2019. Addictive features of social media/messenger platforms and freemium games against the background of psychological and economic theories.International journal of environmental research and public health16, 14 (2019), 2612

    work page 2019

  37. [37]

    Matej Mosnar, Adam Skurla, Branislav Pecher, Matus Tibensky, Jan Jakubcik, Adrian Bindas, Peter Sakalik, and Ivan Srba. 2025. Revisiting Algorithmic Audits of TikTok: Poor Reproducibility and Short-term Validity of Findings. InProceed- ings of the 48th International ACM SIGIR Conference on Research and Development in Information Retrieval(Padua, Italy)(SI...

    work page doi:10.1145/3726302.3730293 2025

  38. [38]

    Ruvimbo Musiyiwa and Jenna Jacobson. 2023. Sponsorship disclosure in social media influencer marketing: The algorithmic and non-algorithmic barriers.Social Media+ Society9, 3 (2023), 20563051231196870. The DSA’s Blind Spot: Algorithmic Audit of Advertising and Minor Profiling on TikTok

    work page 2023

  39. [39]

    Agnes Nairn and Cordelia Fine. 2008. Who’s messing with my mind? The implications of dual-process models for the ethics of advertising to children. International Journal of Advertising27, 3 (2008), 447–470

    work page 2008

  40. [40]

    Jessica Packer, Helen Croker, Anne-Lise Goddings, Emma J Boyland, Claire Stans- field, Simon J Russell, and Russell M Viner. 2022. Advertising and young people’s critical reasoning abilities: Systematic review and meta-analysis.Pediatrics150, 6 (2022), e2022057780

    work page 2022

  41. [41]

    Cecilia Panigutti, Delia Fano Yela, Lorenzo Porcaro, Astrid Bertrand, and Josep Soler Garrido. 2025. How to investigate algorithmic-driven risks in online platforms and search engines? A narrative review through the lens of the EU Digital Services Act. InProceedings of the 2025 ACM Conference on Fairness, Ac- countability, and Transparency (FAccT ’25). As...

    work page doi:10.1145/3715275.3732052 2025

  42. [42]

    Oreste Pollicino. 2023. The quadrangular shape of the geome- try of digital power(s) and the move towards a procedural digi- tal constitutionalism.European Law Journal29, 1-2 (2023), 10–

    work page 2023

  43. [43]

    arXiv:https://onlinelibrary.wiley.com/doi/pdf/10.1111/eulj.12472 doi:10.1111/eulj.12472

    work page doi:10.1111/eulj.12472

  44. [44]

    Manoel Horta Ribeiro, Raphael Ottoni, Robert West, Virgílio A. F. Almeida, and Wagner Meira. 2020. Auditing radicalization pathways on YouTube. In Proceedings of the 2020 Conference on Fairness, Accountability, and Transparency (Barcelona, Spain)(FAT* ’20). Association for Computing Machinery, New York, NY, USA, 131–141. doi:10.1145/3351095.3372879

    work page doi:10.1145/3351095.3372879 2020

  45. [45]

    Jan Ole Rixen, Luca-Maxim Meinhardt, Michael Glöckler, Marius-Lukas Ziegen- bein, Anna Schlothauer, Mark Colley, Enrico Rukzio, and Jan Gugenheimer

    work page

  46. [46]

    The loop and reasons to break it: Investigating infinite scrolling behaviour in social media applications and reasons to stop.Proceedings of the ACM on Human-Computer Interaction7, MHCI (2023), 1–22

    work page 2023

  47. [47]

    Esther Rozendaal, Matthew A Lapierre, Eva A Van Reijmersdal, and Moniek Buijzen. 2011. Reconsidering advertising literacy as a defense against advertising effects.Media psychology14, 4 (2011), 333–354

    work page 2011

  48. [48]

    Christian Sandvig, Kevin Hamilton, Karrie Karahalios, and Cedric Langbort. 2014. Auditing algorithms: Research methods for detecting discrimination on internet platforms.Data and discrimination: converting critical concerns into productive inquiry22 (2014), 4349–4357

    work page 2014

  49. [49]

    Jakub Simko, Matus Tomlein, Branislav Pecher, Robert Moro, Ivan Srba, Elena Ste- fancova, Andrea Hrckova, Michal Kompan, Juraj Podrouzek, and Maria Bielikova

    work page

  50. [50]

    InAdjunct Proceedings of the 29th ACM Conference on User Modeling, Adaptation and Personalization(Utrecht, Nether- lands)(UMAP ’21)

    Towards Continuous Automatic Audits of Social Media Adaptive Behavior and its Role in Misinformation Spreading. InAdjunct Proceedings of the 29th ACM Conference on User Modeling, Adaptation and Personalization(Utrecht, Nether- lands)(UMAP ’21). Association for Computing Machinery, New York, NY, USA, 411–414. doi:10.1145/3450614.3463353

    work page doi:10.1145/3450614.3463353

  51. [51]

    Sara Solarova, Matúš Mesarčík, Branislav Pecher, and Ivan Srba. 2026. Beyond the Checkbox: Strengthening DSA Compliance Through Social Media Algorithmic Auditing. arXiv:2601.18405 [cs.CY] https://arxiv.org/abs/2601.18405 Presented at 2026 CHI Conference on Human Factors in Computing Systems

    work page arXiv 2026

  52. [52]

    Ivan Srba, Robert Moro, Matus Tomlein, Branislav Pecher, Jakub Simko, Elena Stefancova, Michal Kompan, Andrea Hrckova, Juraj Podrouzek, Adrian Gavornik, and Maria Bielikova. 2023. Auditing YouTube’s Recommendation Algorithm for Misinformation Filter Bubbles.ACM Trans. Recomm. Syst.1, 1, Article 6 (Jan. 2023), 33 pages. doi:10.1145/3568392

    work page doi:10.1145/3568392 2023

  53. [53]

    Ivan Srba, Branislav Pecher, Jakub Simko, Robert Moro, and Maria Bielikova

    work page

  54. [54]

    In Proceedings of European Workshop on Algorithmic Fairness, EW AF’25

    Model-based Algorithmic Auditing of Social Media AI Algorithms. In Proceedings of European Workshop on Algorithmic Fairness, EW AF’25. June 30 – July 02, 2025. Eindhoven, NL

    work page 2025

  55. [55]

    Laurence Steinberg. 2004. Risk taking in adolescence: what changes, and why? Annals of the New York Academy of Sciences1021, 1 (2004), 51–58

    work page 2004

  56. [56]

    2025.Branded Content Policy

    TikTok. 2025.Branded Content Policy. TikTok. https://www.tiktok.com/legal/ page/global/bc-policy/en Last updated July 2025

    work page 2025

  57. [57]

    2025.Community Guidelines

    TikTok. 2025.Community Guidelines. TikTok. https://www.tiktok.com/ community-guidelines/en-GB Released April 17, 2024; updated September 13, 2025

    work page 2025

  58. [58]

    2025.TikTok Library FAQ

    TikTok. 2025.TikTok Library FAQ. TikTok. https://library.tiktok.com/faq Frequently Asked Questions about the TikTok Library (accessed online)

    work page 2025

  59. [59]

    TikTok. 2025. TikTok – DSA Transparency Report: January – June 2025. https://sf16-va.tiktokcdn.com/obj/eden-va2/zayvwlY_fjulyhwzuhy%5B/ ljhwZthlaukjlkulzlp/DSA_H1_2025/TikTok-DSATransparencyReport-January- June-2025.pdf. Accessed: 2026-01-07

    work page 2025

  60. [60]

    2025.About the Commercial Content Disclosure Setting for Creators

    TikTok For Business. 2025.About the Commercial Content Disclosure Setting for Creators. TikTok. https://ads.tiktok.com/help/article/about-the-content- disclosure-setting-for-creators Last updated December 2025

    work page 2025

  61. [61]

    Matus Tomlein, Branislav Pecher, Jakub Simko, Ivan Srba, Robert Moro, Elena Ste- fancova, Michal Kompan, Andrea Hrckova, Juraj Podrouzek, and Maria Bielikova

    work page

  62. [62]

    InProceedings of the 15th ACM Conference on Recommender Systems(Amsterdam, Netherlands)(RecSys ’21)

    An Audit of Misinformation Filter Bubbles on YouTube: Bubble Bursting and Recent Behavior Changes. InProceedings of the 15th ACM Conference on Recommender Systems(Amsterdam, Netherlands)(RecSys ’21). Association for Computing Machinery, New York, NY, USA, 1–11. doi:10.1145/3460231.3474241

    work page doi:10.1145/3460231.3474241

  63. [63]

    Aleksandra Urman, Mykola Makhortykh, and Aniko Hannak. 2025. WEIRD Audits? Research Trends, Linguistic and Geographical Disparities in the Algo- rithm Audits of Online Platforms - A Systematic Literature Review. InProceed- ings of the 2025 ACM Conference on Fairness, Accountability, and Transparency (FAccT ’25). Association for Computing Machinery, New Yo...

    work page doi:10.1145/3715275.3732026 2025

  64. [64]

    Eva A Van Reijmersdal, Sophie C Boerman, Moniek Buijzen, and Esther Rozen- daal. 2017. This is advertising! Effects of disclosing television brand placement on adolescents.Journal of youth and adolescence46, 2 (2017), 328–342

    work page 2017

  65. [65]

    Eva A Van Reijmersdal, Esther Rozendaal, Nadia Smink, Guda Van Noort, and Moniek Buijzen. 2017. Processes and effects of targeted online advertising among children.International journal of advertising36, 3 (2017), 396–414

    work page 2017

  66. [66]

    Georgia Wells, Jeff Horwitz, and Deepa Seetharaman. 2021. Facebook Knows Instagram Is Toxic for Teen Girls, Company Documents Show.The Wall Street Journal(14 September 2021). https://https://www.wsj.com/tech/personal- tech/facebook-knows-instagram-is-toxic-for-teen-girls-company-documents- show-11631620739

    work page 2021

  67. [67]

    Pieter Wolters and Frederik Zuiderveen Borgesius. 2025. The EU Digital Services Act: what does it mean for online advertising and adtech?International Journal of Law and Information Technology33 (2025), eaaf004. doi:10.1093/ijlit/eaaf004 A Prompts Utilized for Automatic Data Annotation The prompt used by LLM (GPT-4.1) as a part of theuser interaction pred...

    work page doi:10.1093/ijlit/eaaf004 2025

  68. [68]

    Detection: Determine if the image contains an advertisement

    work page

  69. [69]

    Sponsored

    Type Classification: If it is an ad, classify the TYPE based on visual indicators: - Formal: Look for platform-injected labels "Sponsored" or "Ad" displayed on the bottom of the video. - Influencer: Look for creator-injected labels "Paid partnership" or "Promotional content" displayed on the bottom of the video, - Other: Look for videos with commercial co...

    work page

  70. [70]

    inside your string values. Use single quotes'instead. Example:

    Topic Classification: Classify the primary content of the ad into exactly one of these categories: -`beauty`(makeup, skincare, cosmetics, clearskin, kbeauty, glasskin) -`fitness`(abs, workout, gym, sports, health, nutrition, gymtok , supplements) -`gaming`(video games, consoles, streamers, gamer, gaming, gamerlife) -`politics`(political news, political de...

    work page 2007