The next session I’m attending at the 2026 International Communication Association conference in Cape Town starts with a paper by Simon Kruschinski, whose focus is on political microtargeting in election advertising. This is now a growing practice in elections around the world, and the perception of such ads by their targets depends on the congruence between advertising approaches and the sociodemographics of target groups.
But we are still missing rich qualitative data on the situational perception and evaluation of such ads by users in a realistic social media setting. The present study explored this through a think-aloud protocol study with 12 participants from Germany, representing conservative and Greens supporters; participants logged onto a computer with their own Facebook accounts, a tool then extracted 25 posts from that feed, and mock-up ads were then injected into this feed.
Participants were asked to voice their thoughts as they were doing so, and ChatGPT was then used to analyse the transcripts of this stream-of-consciousness narration, following an iterative process of prompt engineering. This was done over two rounds: an unaided round without prompting by researchers, and an aided round with prompts from the research team.
In the unaided round, few participants responded to the ads at all; in the aided round, participants mostly noticed ads which did not match their own political orientation, while they were occasionally welcoming ads that did align with their own views. Negative responses were largely directed at the parties and their policies, rather than the specific ads themselves. Some users saw such microtargeting as legitimate (since companies do this too); others had significant privacy concerns.
This is only a small-scale study, of course, and EU regulations on political microtargeting have changed recently, so further research is necessary.
