AI skin analysis tools, which claim to offer personalised skincare insights and product recommendations based on selfies, have been appearing at major retailers such as Boots and Superdrug, as well as on beauty brand websites. However, a recent test found that these tools can provide inconsistent results and inaccurate recommendations.
The test, which involved 30 people using the Boots Skincare Diagnostic Tool for over 120 scans, revealed that while the tool gave consistent overall scores in identical lighting, it struggled to detect self-reported localised issues like acne, breakouts, redness, and hyperpigmentation. For instance, out of 14 people with visible acne, the scanner failed to spot it or contradicted them 10 times. Overall, approximately 37% of testers received results that did not match their actual skin concerns.
Good lighting was found to be crucial for accuracy, with scores swinging wildly for about half the group when light sources shifted. Skin tone also played a significant role, with the app struggling to detect hyperpigmentation and dark spots on participants with darker skin tones, often giving them higher overall health scores.
Product recommendations were also identified as a concern. Despite questionnaires, the tools sometimes recommended products that could aggravate existing conditions, such as retinol and salicylic acid for a tester with rosacea.