Comprehending the intricacies of emotions is essential for successful human interaction, where being able to discern someone’s emotional state is key to truly understanding their behavior and consequently, their character [1]. However, emotions can be manipulated, with individuals concealing their true emotions. As a result, this study addressed the detection of deceptive emotions in text using Natural Language Processing (NLP) utilizing Long Short-Term Memory (LSTM). The MSP-IMPROV and MSP-PODCAST datasets, focusing on happiness, sadness, and anger, were employed to assess the model’s effectiveness in scenarios with and without explicit emotional indicators. The findings indicated that with emotion indicators, the model produced an accuracy of 95.29%, compared to a slightly lower accuracy of 89.41% without emotion indicators, in detecting fake emotions in text.