X uses a combination of machine learning models and human review to automatically detect posts that violate its rules, including spam, hateful content, and manipulation campaigns. The system can take action automatically or surface content to human moderators.

According to X's October 2024 DSA Transparency Report, X uses combinations of natural language processing models, image processing models, and other machine learning methods to detect potentially violating content. Both machine learning and heuristic models are trained on data labeled by human content moderators. Automated enforcements undergo testing before going live, and the system either acts automatically or surfaces content to human reviewers based on user reports or proactive detection. X's Global Transparency Report for H1 2024 stated that defenses for manipulation and spam are 'primarily proactive or automated.'