Can artificial intelligence detect skin cancer better than a dermatologist?
This is the headline question that has generated the most coverage in recent years, and the honest answer is: in certain circumstances, yes. Deep learning systems (which learn to recognise patterns by analysing millions of images) have achieved levels of diagnostic accuracy comparable to or surpassing those of many specialists under standard visual inspection conditions.
The most significant regulatory milestone came in January 2024, when the FDA authorised DermaSensor, the first AI-powered medical device designed to detect the three most common types of skin cancer (melanoma, basal cell carcinoma, and squamous cell carcinoma) in the primary care setting. The clinical studies supporting its authorisation showed that the device reduced the rate of undetected cancers from 18% to 9%. Its authorisation established an unprecedented precedent: for the first time, an artificial intelligence device was approved for use by non-specialist physicians, extending the reach of early detection beyond the dermatology clinic.
At the research level, a review published in Bioengineering in 2025 confirms that the most advanced models already exceed the threshold of 98% diagnostic accuracy on international reference datasets, compared to 65–79% for visual inspection without a dermatoscope and 82–91% for specialist-assisted dermoscopy.
How is AI changing the management of inflammatory diseases?
Beyond skin cancer, artificial intelligence is opening new avenues in the management of conditions such as atopic dermatitis, psoriasis, various types of alopecia, and vitiligo. A review published in the Journal of Investigative Dermatology in 2026 by researchers from the University of California identifies the three areas of greatest clinical impact:
- The first is the objective measurement of severity: AI systems can analyse clinical photographs and calculate severity indices with a consistency that surpasses the usual variation between different human observers.
- The second is the prediction of treatment response: prior to initiating a biological therapy, certain models are capable of estimating the probability of response in a specific patient.
- The third is the identification of subgroups of patients who will benefit from specific drugs, opening the door to genuinely precision dermatology.
What are large language models such as ChatGPT used for in dermatology?
The arrival of large language models such as ChatGPT, Gemini, and Claude has added a new dimension to clinical AI. A systematic review published in the Journal of Biomedical Sciences in 2025, which analysed 17 studies, concludes that these systems have matched or surpassed human performance in standardised medical examinations, and demonstrate proven utility in three areas: patient education, the extraction and organisation of information from unstructured clinical records, and support for clinical decision-making.
What can artificial intelligence not do in dermatology?
The scientific literature is clear in identifying the relevant limitations that are worth being aware of. The majority of cancer detection models have been trained on images that underrepresent darker skin phototypes, which compromises their performance in populations with darker skin tones. Validation studies under real clinical conditions (outside the controlled laboratory setting) are still scarce. And no artificial intelligence system has to date demonstrated an improvement in clinical outcomes in large-scale controlled prospective trials.
AI is today a highly valuable support tool, but it is by no means a substitute for dermatological judgement.
What no algorithm can replace
Dermatological diagnosis integrates information that goes beyond the image: the patient’s context, their clinical history, the evolution of a lesion over time, and the clinical judgement accumulated over years of practice. Artificial intelligence will, in the coming years, become a first-rate ally for the dermatologist (reducing errors, extending the reach of early detection, and personalising treatments). However, the role of the specialist as the centre of the clinical relationship remains an essential part of the doctor-patient relationship.
If you would like to know how technology and artificial intelligence can improve the diagnosis and treatment of your condition, at our clinics we integrate the tools with the greatest scientific evidence alongside the expertise of specialist physicians.
