Scientific objectification: heralding precision cosmetics

The beauty industry is undergoing profound change, towards more predictive, evidence-based cosmetics. To meet the demands of consumers and strict regulatory requirements, industry players are using the latest AI-based evaluation methods to demonstrate formulation efficacy with ever-greater precision, and to support claims through objective scientific validation.

By

Ulysse Lardy

In a rapidly evolving beauty industry, AI now offers a cross-cutting impetus capable of orchestrating and analysing vast data streams, thus generating novel scientific objectification techniques that go beyond mere surface texture analysis.

By 2026, AI-based evaluation will be applied comprehensively to sensory, emotional, instrumental and biological factors, to ensure all claims are scientifically validated.

From sensory evaluation to emotional interpretation

Widely used in the consumer goods industry, sensory evaluation involves measuring how a formulation is perceived via the human senses of trained panellists. Texture, scent, appearance and post-application feel are translated into quantifiable attributes, e.g. fresh, smooth or silky feel; fragrant notes or a pearlescent effect. Conversely, consumer tests use untrained panellists to assess whether they like a product. The challenge for brands is to decode the sensory or pleasurable ʺfeelingʺ and then convert it into data for statistical and predictive use.

Driven by the rise of neurocosmetics and a more holistic approach to beauty, sensory studies now analyse the emotional aspect of the product experience by measuring the emotions evoked upon application, i.e. pleasure, sadness, fear, disgust, surprise and anger. Using biometric tools, these new attributes are assessed in terms of their behavioural, physiological or cognitive aspects. At L’Oréal R&I, Wang et al. have used EEG to assess the emotions experienced when wearing face masks. At Shiseido, drawing on their work in brain imaging methods, researchers have used MRI and AI to identify the emotional markers of stress, thus paving the way for innovative products: an ʺanti-stress odourʺ body mist in 2025 and, more recently, a de-stressing sun mist.

All these biometric measurements are performed by specialist companies such as Eurofins, SGS Proderm or Intertek – their main benefit being their ability to reduce cognitive biases and language barriers among panellists, thus significantly improving data robustness. The primary constraint on neurocosmetics development, though, is regulatory: given that cosmetics need to retain their status as beauty or personal care products, under no circumstances can they claim ʺneurologicalʺ effects in the therapeutic sense. A day cream cannot claim to 'enhance well-being' or ʺreduce stressʺ upon application. Currently, each lab uses its own, separate biometric techniques. One possible solution for regulating these practices could, therefore, involve harmonising testing protocols around standards such as ISO for suncare products. Regulatory authorities will also need to monitor brand messaging to prevent misleading advertising and ʺneuro-washingʺ.

Alongside these developments, AI is gradually changing the way in which the collected data is used. Recent research, e.g. He et al. on machine learning, demonstrates that sensory data can be correlated with the physicochemical properties of formulations. Predictive algorithms can now simulate an emulsion's sensory profile upstream, prior to manufacture, thus making it possible to determine whether to proceed with the test. Eventually, these algorithms could possibly provide a product's emotional profile, even prior to its manufacture.

Redefining instrumental objectification

Instrumental evaluation techniques are also changing rapidly. Traditionally used for the physicochemical analysis of formulations or the ʺjus” (viscosimeter, pH meter, etc.), the advent of biometrology tools such as the texturometer (blush adhesiveness), rheometer (foundation spreadability), gloss meter (lip gloss shine), etc, means they can now also be used for more in-depth analysis of product behaviour. Tribology offers a more detailed analysis of lubrication and friction phenomena, such as in a body cream or shampoo. The Shiseido group has developed two high-precision sensors: one for contact force, the other for friction during application of the cosmetic to the skin. Likewise, Microphotonics (with its Tribotouch model) has incorporated biomimetic applicators into its machines to simulate a finger pressing on the skin. In makeup, goniophotometry (e.g. Bossanova's Samba model) is used to identify the source of the shine in a lip gloss or lipstick. Lastly, in perfumery, AI-controlled electronic noses detect and classify odour molecules objectively and continuously.

Rapid advances are also being made in the analysis of dispersed systems, with labs now using dynamic light scattering to accurately predict incompatibilities or instabilities by measuring droplet size in an emulsion, or dispersion in sunscreens. Another example is hyperspectral imaging, which is opening up new possibilities for assessing pigment distribution in tinted formulations. Stability and compatibility studies are usually conducted macroscopically in ovens, supplemented by accelerated ageing techniques (centrifuge, sun test). It takes time, though, to collect organoleptic, physicochemical and microbiological data on the product, meaning that it can take several months to detect an instability, thus requiring reformulation and disrupting development. To address this issue, companies such as Fluigent or Microtrac are using microfluidics to assess product stability microscopically. Instead of creating a conventional white emulsion in a beaker, water and oil droplets of the same size are injected onto a chip (organ-on-a-chip). A set of cameras, pressure sensors, laser probes and microelectrodes, coordinated by AI, enables the immediate detection of any form of instability.

In terms of product safety studies, microbiological testing is becoming increasingly automated. The Korean subcontractor, Kolmar, recently developed an alternative challenge test method based on AI and robotics. The company claims that this method reliably produces results 2.5 times faster than in a conventional lab. Also in Korea, researchers at AmorePacific have developed an initial skin irritation test protocol based on an automated reading algorithm.

From state-of-the-art ex vivo research…

Research in ex vivo testing is accelerating, driven by the ban on in vivo animal testing and the relaxation of regulations regarding alternative methods in certain countries (such as the US). As such, 2026 marks the advent of ʺaugmentedʺ skin models derived from bioprinting, live explants or organoids.

Bioprinting is a manufacturing method that involves the layer-by-layer deposition of ʺbio-inksʺ (a mixture of live cells and biomaterials) based on a digital model. The technique now also covers vascularisation (penetration of actives into various strata at L’Oréal), and pigmentation (efficacy of suncare and anti-dark spot products by LabSkin Creations). Sophisticated learning algorithms and predictive models are used to simulate and control the organisation of these multilayered structures in real time. Next-generation ex vivo approaches are also focusing on the skin microbiome, one example being Genoskin, which uses explants of live human skin grown on a nutrient matrix to support the testing of products that claim to be ʺmicrobiome-friendlyʺ.

Organoids are more complex systems in which pluripotent stem cells are cultured and multiply. They may, for example, feature the ECM and its structural components (such as GAGs), or even other appendages such as hair. Automated data processing and computational modelling provide insight into specific biological mechanisms, such as skin ageing or hair greying. These advances will help drive the development of high-precision bioactive hair dyes and anti-ageing skincare products.

At the 2025 IFSCC, CTI Biotech unveiled an innervated system that will be used to analyse sensory responses or the efficacy of products for sensitive skin.

… to real-time in vivo research

Clinical research is also benefiting from increasingly sophisticated, non-invasive, ultra-fast measurement and analysis tools. One major breakthrough in 2026 is the use of high-speed atomic force microscopy (HS-AFM), which takes just 5 seconds to produce a high-resolution image of the nano-textured surface of corneocytes (stratum corneum cells). This technique could offer new ways of objectively assessing the smoothing or moisturising effect of a skincare product at nanoscale. AI- based computer vision algorithms excel at this task, instantly processing massive volumes of high-resolution images and automating the detection of micro-changes in skin texture.

The most striking development, though, is probably the way clinical assessments are gradually shifting from conventional research centres, towards an approach more anchored in daily life. In addition to smartphone sensors, volunteers are fitted with smart patches that measure skin hydration, cortisol levels or UV exposure in real time over a period of several weeks. Once again, the use of AI dominates here, essential to centralising and correlating these continuous streams of big data. This approach enriches clinical studies with behavioural and environmental data that are far more representative of the products' actual use. No longer limited to a one-off snapshot taken in the lab, cosmetic evaluation is now moving towards continuous monitoring of the interaction between the formula, the skin and the consumer’s lifestyle.

Towards bespoke, real-world cosmetics

These advances mean that the cosmetics industry is no longer restricted to, at times, extravagant or unfounded claims supported by tests that are often poorly understood by the general public. The combined effect of regulatory requirements, growing access to scientific information and technological breakthroughs is driving the industry to redefine itself around a set of holistic, objective methods. Consumers are no longer considered just in terms of their skin type (normal, oily, phototype, etc.). These new methods take account both of the product and of the user’s actual experience, in terms of physical and behavioural characteristics and cognitive abilities. Product profiles are now directly enhanced via new subjective, physicochemical and biological parameters, derived by combining results from advanced ex vivo models (bioprinting, organoids), emotional (EEG) and biometrological (tribology, goniophotometry) measurements and increasingly sophisticated in vivo clinical trials. Indirectly, ex vivo models further our understanding of the skin and its mechanisms, thereby enabling the development of more sophisticated cosmetics – with AI emerging as the most effective means of orchestrating this wealth of ultra-complex data. The convergence of neuroscience with formulation, tissue engineering, biophysics and robotics is already opening up avenues for entirely new categories and novel claims.

These methods could also be applied very effectively to the field of packaging (e.g. emotional impact of a perfume bottle). Broadly speaking, these changes are gradually bringing cosmetics closer to disciplines like dermatology, neurology and dermopharmacology. A convergence that is reshaping not just the way the beauty industry conducts product evaluations, but also the way it frames its approach to innovation.