According to the World Health Organization, anxiety is the most common emotional disorder and is often accompanied by high comorbidity. Our study investigates using social media posts to analyze emotional dynamics related to anxiety. It examines 1000 anxious and 1000 non-anxious Reddit users, applying XGBoost ensemble technique and comparing it to decision tree, random forest (RF), and KNN models. The RF model achieved the highest accuracy (88%) and better F1-scores for both classes, outperforming XGBoost. However, XGBoost showed slightly lower accuracy and struggled more with classifying the minority class (non-anxiety), indicating its limitations in distinguishing between anxious and non-anxious users.