Introduction: Artificial intelligence (AI) is transitioning from experimental image classification to clinical deployment in dermatology. However, critical gaps remain regarding real-world performance, regulatory maturity, and equity across skin types. Methods: This systematic review synthesized data from 80 primary studies (2017–2026) evaluating AI applications in clinical dermatology, including prospective trials, randomised controlled trials (RCTs), and real-world deployment studies. Outcomes included diagnostic accuracy, clinician performance changes, healthcare efficiency, and subgroup equity. Results: In meta-analyses, deep learning achieved pooled sensitivity of 82% and AUC of 0.92 for melanoma detection from dermoscopic images, matching or exceeding dermatologists (p<0.01) (17,22,37). In prospective NHS deployment (25,788 lesions), an AI medical device triaged 98.6% of high-risk skin cancers into urgent pathways, significantly outperforming teledermatologists (95.9%, p=0.004) (3). AI assistance improved non-specialist sensitivity by +27.9 percentage points but only +2.1 points for experts (6). The only consumer AI RCT (n=19,009) showed no cancer detection benefit but increased benign lesion claims (p<0.001) and costs (€63 vs €47, p<0.001) (7). AI systems performed significantly worse on Fitzpatrick IV–VI skin (AUROC 0.82 vs 0.89 for I–III, p<0.01) (1). For inflammatory diseases, AI severity assessment achieved pooled sensitivity 80.5% and specificity 96.2% (2). Discussion: The evidence most strongly supports UKCA/CE-approved AI within specialist teledermatology triage for urgent skin cancer pathways, demonstrating high sensitivity and efficiency gains. AI benefit is expertise-dependent, greatest for non-specialists. Consumer-facing AI currently increases costs without proven population benefit. Substantial skin-type performance inequities require urgent correction. Conclusion: AI is clinically useful for skin cancer triage in specialist pathways but not ready for autonomous consumer screening. Prospective validation across diverse populations and regulatory mandates for equity reporting are necessary