Acne vulgaris is a common skin problem that affects many people around the world, especially teenagers and young adults. Conventional diagnosis methods are largely manual and subjective, often resulting in inconsistent outcomes. To address these challenges, we propose AcneMultiNet, a hybrid deep learning model that integrates the complementary strengths of VGG16 and MobileNetV2 architectures for accurate and efficient acne detection. The model employs feature-level fusion to extract and combine discriminative representations from both networks, enhancing classification performance while maintaining computational efficiency. Trained on a dataset comprising 14,179 labeled facial images, the proposed model achieves a test accuracy of 99.96%, outperforming several baseline convolutional neural networks. Quantitative evaluations based on precision, recall, F1-score, and ROC confirm the model’s robustness and its suitability for clinical and mobile applications. The results highlight AcneMultiNet’s potential as a practical and scalable solution for automated acne diagnosis.