What visual misinformation appears in social media feeds when accounts follow partisan actors during elections? This presentation introduces some preliminary findings and ongoing research objectives for a study conducted across four national elections in 2024 in Belgium, France, the United Kingdom and the United States. For the study, two dummy accounts were set up, each on four social media platforms: Twitter / X, Facebook, Instagram and TikTok for each country. Each account would either follow influential right-wing or left-wing accounts, with the aim of training the algorithm to reveal visual misinformation that users might be encountering on the platform.

Visual misinformation that appeared on their feeds was gathered and archived. The first analysis of the data gathered involved a content analysis, comparing country-specific dataset as well as looking for evidence of visual disinformation created by artificial intelligence image generators. The findings in this regard are somewhat surprising. AI does play a role, but other forms of visual misinformation are also prominent in the dataset. Furthermore, the aim of AI generated visual misinformation is not always to mislead, but rather produce content that is obviously not real, but plays into well-established and false political narratives and talking points. The presentation will also incorporate the preliminary results of a set of interviews with experts in each of the four countries.

The next EDMO BELUX Lunch Lecture will take place on September 15th, 2025, at 12:00-13:00 CEST. During the webinar, where Nick Anstead and Bart Cammaerts will present preliminary findings of a study on exposure to visual misinformation during recent elections in several countries.

Attendance is free but registration is required here.