The study aims to develop an audio deepfake detection model integrating both acoustic and textual features for the limitations of the methods compared in traditional detection and to enhance the reliability of identifying spoofed audios. We used a multi-view approach, combining acoustic features such as Mel Frequency Cepstral Coefficients (MFCC), Linear Frequency Cepstral Coefficients (LFCC), and pitch analysis with textual embed-dings generated by OpenAI’s text-embedding models. Experiments were conducted using the ASVspoof 2019 dataset, and the proposed models were evaluated using machine learning algorithms, including Random Forest, XGBoost, and LightGBM. The results show that the textual features significantly complement the acoustic analysis, giving better accuracy and robustness in the detection compared to the single-view classification models. The best performance was by LightGBM with embedded features: EER of 13.62%, accuracy of 83.80%, precision of 98.57%, recall of 83.14%, and F1-score of 90.20%. These findings show the efficiency of the multi-view learning approach for the enhancement of audio deepfake detection.