This paper explores clickbait detection using Transformer models, specifically IndoBERT and RoBERTa. The objective is to leverage the models specifically for clickbait detection accuracy by employing balancing and augmentation techniques on the dataset. The research demonstrates the benefit of balancing techniques in improving model performance. Additionally, data augmentation techniques also improved the performance of RoBERTa. However, it resulted differently for IndoBERT with slightly decreased performance. These findings underline the importance of considering model selection and dataset characteristics when applying augmentation. Based on the result, IndoBERT, with a balanced distribution, outperformed the previous study and the other models used in this research. Furthermore, by incorporating balancing and augmentation techniques, the research surpasses previous studies, contributing to the advancement of clickbait detection accuracy. This work highlights the value of leveraging pre-trained Transformer models and specific dataset-handling techniques. The implications include the necessity of dataset balancing for accurate detection and the varying impact of augmentation on different models. These insights aid researchers and practitioners in making informed decisions for clickbait detection tasks, benefiting content moderation, online user experience, and information reliability. The study emphasizes the significance of utilizing state-of-the-art models and tailored approaches to improve clickbait detection performance.