The research on conspiracy beliefs suffers from the lack of longitudinal data. To fill this gap, we use a merged data set from the 2021 parliamentary and 2023 presidential election online panels (with around 750 respondents participating in both panels) to measure and explain the stability of their conspiracy belief. Besides the survey data, we use web-tracking data, offering a granular understanding of participants' media consumption patterns (including their consumption of content on so-called disinformation media). By linking respondents' conspiracy beliefs with their online information exposure during the 2021 and 2023 election campaigns, we aim to identify the role of digital media in either reinforcing or challenging conspiracy belief.