The Limitations of AI-Powered Skin Analysis in Retail Stores
Why This Is Being Discussed Now
In mid-October 2026, the British consumer association *Which?* published an investigation into an AI-based skin analysis tool offered by a major UK retail chain, involving thirty individuals and over 120 facial scans. The findings are relevant to suppliers of pharmacies and large retail chains, as they highlight areas where this technology remains flawed.
First Limitation: Results Depend on Lighting
Under identical lighting and distance conditions, repeated scans yielded nearly consistent scores. However, when the lighting was altered, scores for about half of the participants fluctuated—increasing by as much as 39 points or decreasing by up to 25. Since retail environments vary—featuring different lighting, mirrors, and window placements—the same skin can receive vastly different assessments depending on the specific store or time of day.
Second Limitation: Accuracy and Skin Tones
Among the fourteen participants who identified visible acne as their primary concern, the tool either failed to detect it or provided a contradictory assessment in ten cases. Six out of nine participants with dark spots were assessed as having healthy pigmentation, and none of the seven individuals with redness had it flagged by the system. Overall, only about 63% of the results aligned with the participants’ own observations, and hyperpigmentation proved more difficult to detect on skin phototypes IV and V.
Third Limitation: Recommendations Are Not Neutral
When participants changed their responses to the questionnaire provided before the scan, twenty-five out of thirty received an entirely different list of product recommendations. Certain brands appeared far more frequently than others; in one instance, a person with rosacea was even recommended products containing retinol and salicylic acid—ingredients that can be potentially irritating. The dermatologist consulted by *Which?*… It was noted that these tools do not constitute a medical diagnosis and that terms such as “radiance” or “opacity” are not recognized in clinical dermatology.
The hidden cost: privacy and compliance
Scanning a customer’s face involves processing personal images; this requires a privacy notice and a legal basis for data processing. Furthermore, the retail outlet providing the tool also bears responsibility. As of August 2, 2026, the transparency obligations of the EU AI Act will also apply, including the requirement to inform individuals when they are interacting with an AI system. For a retailer, this entails operational costs and the risk of reputational damage should the results prove inaccurate.
Physical cosmetic testers: what they measure and why they are easier to manage
The IPS cosmetic tester is a 3-in-1, single-use kit. The “Skin Test” identifies skin type (dry, oily, or normal); the “Hydration Test” measures hydration levels on the face and body; and the “Wrinkle Test” provides a visual assessment of wrinkles and signs of aging. The test can be performed independently and privately in just a few seconds—no app, no account, and no collection of images or personal data required. The product is non-toxic, “Made in Italy,” and available for private labeling with custom packaging; it can be supplied as a complete kit or as individual testers.
The limitations of physical testers
To be candid: these tools measure only three parameters; they do not replace a professional dermatological opinion, they do not create a personalized skincare routine, and they do not maintain a historical record over time. As they are single-use, they require constant restocking. Their value lies in their simplicity: they provide an immediate indication that guides the purchasing decision and can serve as a useful supplement to a pharmacist’s professional advice.
What a buyer should consider
The decision should not be based on innovation alone, but rather on reliability, operational costs, and liability. An AI-based system should only be offered to the public following independent validation—tested across various lighting conditions and skin tones—and with a clear framework regarding how data is handled. A physical tester is ordered, displayed, and restocked just like any other product on the shelf.
Frequently Asked Questions
Is AI-based skin analysis reliable for use in a retail setting?
Not yet consistently. In a *Which?* investigation from October 2026, approximately 63% of the results matched what the participants observed; scores fluctuated depending on lighting, and hyperpigmentation proved more difficult to detect on skin phototypes IV and V. It is not a diagnostic tool.
Physical skin tester or AI-based skin analysis: which should a pharmacy choose?
It depends on your objective. A physical tester provides an immediate indication—requiring no app or data collection—and functions as a standard shelf product. AI-based analysis offers the promise of greater personalization, but it requires validation across various lighting conditions and skin tones, as well as the provision of privacy notices to customers and the management of data.
What obligations arise from using AI-based skin analysis in a store?
Because the process involves facial imagery, you must provide a privacy notice and establish a legal basis for data protection. Furthermore, as of August 2, 2026, the transparency requirements of the AI Act apply—including the obligation to inform individuals when they are interacting with an AI system. It is advisable to consult with an expert regarding your specific situation.
Can the IPS cosmetic tester be customized as a private-label product?
Yes. The 3-in-1 single-use kit is available with custom packaging featuring the client’s logo and graphics; it can be provided as a complete set or Individual items in a dedicated box.
Contact us