This work investigates the effectiveness of incorporating paralinguistic feature extraction in audio deepfake detection models. The proposed model extracts paralinguistic features from audio clips and represents them as 1024-dimensional vector embeddings. These embeddings are then used as input for a logistic regression model, which performs binary classification to distinguish between real and deepfake audio samples. The ASVspoof2019 dataset, comprising both genuine and spoofed audio clips, is used to evaluate the model's performance. The results are assessed using evaluation metrics such as Equal Error Rate (EER) and accuracy, which provide insight into the model's effectiveness compared to state-of-the-art methods. The proposed model achieves an EER of 3.04% and an accuracy of 97.9%, indicating that paralinguistic feature extraction is a promising approach for audio deepfake detection. These results suggest that incorporating paralinguistic features can improve the performance of audio deepfake detection systems, making it a valuable tool for future research in this area. Overall, the study demonstrates the potential of paralinguistic feature extraction in enhancing the accuracy and reliability of audio deepfake detection methods.