In this paper, we present an integrated approach for detecting persuasive elements in political advertising. We develop a lightweight model that achieves state-of-the-art performance on Subtask 3 of SemEval 2023 Task 3, and demonstrate its practical application on the Australian Federal Election 2022 Facebook Ads dataset (APA22). We achieve high performance with less computational resources and training data than existing methods, and fine-tune the model with social media content from major news stories and apply it to the APA22 dataset, revealing patterns in how political campaigns use persuasion through funding strategies, word choice, demographic targeting, and changing persuasive strength as election day approaches. The results highlight the need for domain-specific modeling for persuasive analytics in social media, and the potential of uncovering such strategies to increase transparency, inform voters, and increase accountability in digital campaigns.
