In the first ten days of monitoring, which began on 1 March, the researchers found 192 posts containing AI-generated images. These range from realistic deepfakes to stylised graphics. Approximately 90 percent of this content was published without any form of label or disclosure indicating AI involvement. The content covers various topics, from immigration to personal profiling of candidates.

Political communication researcher Fabio Votta, who coordinates the project, notes that the PVV is once again the largest producer of AI campaign content, as was the case during national elections. The tracker shows that AI use occurs in virtually all municipalities studied. Lelystad stands out with 46 AI items, of which 37 originate from PVV candidates. Steenbergen follows with 12 items, all from one local party posting so-called AI slop content.

The researchers point out that generative AI models are becoming increasingly proficient at producing realistic images, making it important for voters to know whether images are real or not. The EU AI Act, which comes into full effect in August, mandates the labelling of AI-generated content. However, at present, the vast majority of AI campaign material contains no disclosure whatsoever. In addition to the tracker, the researchers have developed an interactive tool allowing users to test whether they can recognise AI-generated campaign images.