CampAIgn Tracker
About Us
We are an interdisciplinary team volunteering to ensure transparency about AI usage in political discourse on social media. We met at the Politechathon organized by the Baden-Württemberg Foundation in Berlin, where we won prize money in the “Media Literacy” category, enabling us to further develop the CampAIgn Tracker.
Our team consists of two researchers - Dr. Simon Kruschinski (University of Mainz) and Dr. Fabio Votta (AlgoSoc Consortium at the University of Amsterdam) - who research political campaigns on social media. They brought the initial idea to the group and are responsible for seamless coding of images/videos and their rapid processing on our platform. Miriam Runde, our data scientist, handles backend development and AI detector implementation. Jakob Scherer (background in business administration and IT) and Theresa Schültken (background in politics and strategic foresight) manage external communications and content development.
From left to right in the picture: Theresa Gröninger (Politechathon jury member), Theresa Schültken, Miriam Runde, Fabio Votta, Simon Kruschinski, Jakob Scherer [Credits: @BW Stiftung / Viktor Heekeren]
Methodology
The CampAIgn Tracker helps citizens and journalists understand where AI is being used and in what context through our two-stage process for reliably identifying and coding AI-generated images.
Data Collection
We collect all unpaid posts and paid advertisements from over 3,000 accounts belonging to political parties and leading candidates from SPD, CDU/CSU, FDP, Greens, Left Party, AfD, BSW, Freie Wähler, Animal Protection Party, Progress Party, dieBasis, VOLT, Pirates, Bavaria Party, V-Party, and NPD/Heimat (along with many other minor parties).
We monitor party accounts at federal, state, and local levels:
- Organic posts collected via Brightdata
- Advertisements sourced from Meta Ad Library API
Currently tracking 12070 posts between 2025-01-26 and 2025-02-02.
Account classification is based on the dedicated volunteer work of Who Targets Me, an organization advocating for transparency in digital campaigning. They track online advertising expenditures from over 100,000 advertisers across 650+ parties in more than 50 countries.
AI Detection
Our AI detection is a two-stage process combining automatic classification with AI models and manual annotation:
- Automatic Screening:
- All collected images and video stills are checked for AI generation using the AI detection model from SightEngine.
- Human Validation:
- If the classification probability exceeds 0.1, trained coders manually review the image.
- This ensures automated AI classification is verified through human expertise.
- If the classification probability exceeds 0.1, trained coders manually review the image.
For all identified AI content, coders additionally classify these features:
- Disclosure of AI usage
- Persons depicted in image/video
- Image/video theme
- Actors mentioned in the entire post
- Use of negativity, e.g., through attacks on parties
- Use of acclaims, e.g., through references to party achievements
These comprehensive insights are available through intuitive visualizations on our platform campaigntracker.de.
This enables citizens and journalists to understand how AI is used in campaigns and facilitates public debate about AI’s role in politics.
