Psoriasis is a chronic autoimmune human disorder that affects humans both physically and mentally over a period. Early detection is a challenging task for better management and prevention. Given this, this paper demonstrated an AI-enabled approach for automated psoriasis detection using digital imaging. In this retrospective case–control study, a dataset of 150 psoriasis images (Group I) and 150 normal skin images (Group II) were considered where several image patches (420 psoriasis and 310 normal skin) were generated from the captured image frames for the model optimization purpose. Five deep transfer learning frameworks (VGG16, VGG19, Inception, Exception, and ResNet50) were explored for psoriasis classification, utilizing convolutional layers with various filters to detect psoriasis features. The models were trained and tested with default parameters (Optimizer—Adam, learning rate—0.1%, Loss function—Binary cross entropy). The Inception model outperformed others, further improved with fine-tuning using the AdaGrad optimizer. Performance metrics including sensitivity, specificity, and overall accuracy were evaluated, demonstrating the system’s effectiveness in diagnosing psoriasis from clinical images. The Inception model demonstrated superior performance in automated psoriasis detection, initially achieving 89.84% sensitivity, 90.37% specificity, and 92.24% overall accuracy. After fine-tuning the model hyperparameters with the AdaGrad optimizer, the performance was improved to 93.51% sensitivity, 96.42% specificity, and 99.01% overall accuracy, respectively. The fine-tuned transfer learning model with its higher accuracy can be a potential tool for point-of-care psoriasis detection using clinical photographs.