This is a single-arm prospective diagnostic accuracy study designed to clinically validate an artificial intelligence–enabled wound imaging mobile application for the classification of chronic ulcers as healing or non-healing. The application uses a convolutional neural network trained on annotated wound images to extract visual features including colour distribution, surface texture, and tissue morphology, and generates a binary classification output — healing or non-healing — accompanied by a confidence percentage score. The model does not perform geometric wound measurement or tissue segmentation. For each enrolled patient, the examining clinician first records a structured assessment of wound status using pre-defined criteria based on wound size trajectory, granulation tissue quality, infection status, and wound bed appearance. This clinician assessment serves as the reference standard. A standardized wound photograph is then captured using the AI application, which generates its classification independently. Both outputs are recorded separately. The primary objective is to determine the sensitivity, specificity, positive predictive value, negative predictive value, overall accuracy, and Cohen’s kappa of the application against the clinician reference standard. Receiver operating characteristic curve analysis will be performed using the confidence percentage score as a continuous variable to compute the area under the curve. The study population consists of adult patients aged 18 years and above with clinically diagnosed chronic ulcers including diabetic foot ulcers, venous ulcers, and pressure ulcers, presenting to the Department of General Surgery, Saveetha Medical College and Hospital, Chennai. Target sample size is 120 patients. The study is registered prospectively and has received Institutional Ethics Committee approval. No therapeutic intervention is involved. Standard clinical care is not altered by participation.