The rapid growth of user-generated content on social media platforms has made traditional moderation approaches, such as manual review and keyword-based filtering, increasingly ineffective and error-prone. To overcome these limitations, recent research has focused on automated content moderation techniques based on natural language processing and computer vision. Transformer-based models, including BERT and RoBERTa, enable deeper contextual and multilingual understanding of harmful text, significantly improving moderation accuracy while reducing false positives. For visual content, deep learning models such as convolutional neural networks and vision transformers support real-time detection of violent and inappropriate material. Furthermore, multimodal approaches that combine textual and visual information provide improved detection of complex toxic behavior. Recent advances in large language models further support adaptive and scalable moderation systems aligned with evolving online communities. Collectively, these approaches highlight the need for efficient, explainable, and scalable moderation frameworks to ensure safer digital environments.