A method is presented for detecting stock market manipulation using sentiment analysis of posts on X. Utilizing FinBERT for financial sentiment classification and LSTM autoencoders for anomaly detection, the model identifies unusual sentiment patterns indicative of manipulative trading behavior. Experimental results demonstrate the model’s effectiveness in detecting suspicious activities, outperforming baseline methods in accuracy and robustness. This research highlights the importance of social media sentiment in maintaining market integrity and supports the development of real-time detection tools for investor protection and regulatory oversight.