This study examines whether artificial intelligence can effectively distinguish between cutaneous squamous cell carcinoma (cSCC) and common warts/verruca vulgaris (VV) using hematoxylin and eosin stained whole slides. The researchers developed a deep learning model using clustering-constrained attention multiple instance learning (CLAM) to classify H&E-stained slides. The final dataset included 289 slides (148 cSCC, 141 VV), which were consensus diagnosed by dermatopathologists. The AI model achieved an AUROC of 0.96, with 82.4% of cSCC and 97.4% of VV predictions being correct. Attention heatmaps illustrated the model’s focus on histologically meaningful regions for making its decisions. In cSCC, features such as atypical keratinocytes were highlighted, while in VV, features like papillomatosis were of importance. Overall, the results indicate that AI can achieve comparable performances comparable to experts, serving as a valuable supplement for dermatopathologists in challenging cases.