A Comprehensive Study of Bot Detection in Twitter: Evolution, Comparisons, and Challenges

The rise of automated social bots on Twitter threatens online discourse by spreading misinformation, amplifying propaganda, and manipulating public opinion. This review categorizes bot detection methods into Machine Learning (ML), Deep Learning (DL), Graph-based, Transformer-based, and Hybrid approaches. Traditional ML models, such as Decision Trees and SVMs, struggle with evolving bot behaviours, while DL techniques, including BiLSTMs, CNNs, and Transformers, capture complex patterns in text data. Pre-trained Language Models (PLMs) like BERT and RoBERTa further enhance detection through contextualized representations. Graph-based models analyse user interactions to identify bot clusters, and hybrid approaches improve robustness by integrating multiple methodologies. Moreover, This study presents benchmark dataset in this field along with the future directions. This study synthe-sizes research from 2019–2024, identifying challenges such as data imbalance and proposes future directions for advancing scalable, real-time bot detection frameworks.