da Silva, M. I., de Oliveira, L., de Melo, P. O. V., Goga, O., & Benevenuto, F. (2026). Identifying potentially irregular electoral ads in Facebook during the Brazilian elections. ACM Transactions on the Web. https://doi.org/10.1145/3796545
Summary
This paper presents an independent auditing system built to monitor political advertising on Facebook/Meta during the 2018 Brazilian elections. Motivated by concerns about targeted-ad abuse of the kind observed in the 2016 U.S. presidential election, the authors adapted a browser plugin — installed by over 2,000 volunteers — to collect the ads appearing on real users’ timelines and analyze them for signs of irregular electoral advertising. The central argument is that crowdsourced, volunteer-based auditing is a feasible and valuable method for bringing transparency to opaque ad-targeting systems, especially in contexts where platforms fail to provide adequate disclosure. The work extends post-2016 debates on platform accountability from the U.S. into the Brazilian electoral environment.
Key Contributions
- Design and deployment of an independent ad-auditing infrastructure tailored to the Brazilian electoral context.
- Empirical evidence on political advertising practices on Meta during the 2018 Brazilian elections.
- A methodological template for crowdsourced platform auditing that can be applied to other electoral settings.
Methods
- Adaptation of a browser plugin to scrape advertisements appearing on volunteers’ Facebook timelines.
- Recruitment and deployment of the tool to more than 2,000 volunteer users across Brazil.
- Collection of political ads served during the 2018 Brazilian election period.
- Identification and classification of ads as potentially irregular under Brazilian electoral rules.
Findings
- The system successfully gathered a corpus of political ads from real Facebook users during the election.
- A subset of the collected ads were flagged as potentially irregular under Brazilian electoral law.
- Volunteer-based crowdsourcing proved viable as a transparency mechanism where platform-provided disclosure was inadequate.
Connections
This work sits within a growing body of research using independent, crowdsourced, or infrastructural approaches to audit opaque platform advertising and data-access limitations — see Bouchaud2026-lr and Bouchaud2026-np on independent measurement of platform ad systems, and Rieder2025-ju and Rieder2026-pp on the politics of platform data access. Its focus on Brazilian elections and coordinated/irregular political messaging connects it to Latin American election studies such as Ventura2025-sw and Ventura2026-yc, as well as broader work on coordinated and problematic electoral behavior on Meta like Giglietto2020-6278a4aa and Kansaon2025-id.
Podcast
A research-radio episode discusses this paper: 🎧 MP3 · Spotify · Apple Podcasts