Instead of relying on repeated rounds of tweaking, cosmetic brands are starting to treat formulations like engineered systems, using AI and molecular simulation to model, stress-test, and refine them before a single beaker is lifted or a lab sample is made.
Image Credi: Aleksandar Malivuk/Shutterstock.com
As Rudy Coquet, VP Sales, Americas & Europe at Matlantis, puts it, “Beauty R&D has entered a ‘materials-intelligence’ era,” one in which “AI-driven atomic-level simulations are transforming how cosmetic formulations are designed and optimized.”
The promise is straightforward: Run a larger proportion of the trial phase computationally, cut waste, and bake in performance, safety, and regulatory checks before any ingredients reach the bench.
Get all the details: Grab your PDF here!
From Trial-and-error to Atom-level Design
Traditional cosmetic R&D has relied on empirical iteration, in which chemists blended surfactants, emollients, polymers, and actives, then tested stability, texture, and efficacy in cycles that usually lasted months.
While this method ensures quality, it makes it difficult to explore new formulations or understand the interactions between ingredients. As a result, formulators often reused known ingredient “families,” limiting innovation and restricting sustainable options.1
Coquet’s description cuts straight to the point: “A 'materials intelligence' approach brings cosmetics out of time-consuming trial-and-error processes,” shifting work away from endless loops and toward more intentional testing.1
Materials intelligence reframes each formulation as a complex soft-matter system, where properties emerge from atomic-scale interactions among oils, water, polymers, pigments, and biological interfaces such as skin or hair. Similarly, molecular modelling methods like molecular dynamics (MD) and docking connect structure and function by analyzing how molecules assemble, diffuse, bind, and respond to temperature or shear.1
For Coquet, the hinge is prediction: “By predicting ingredient interactions and material properties at the atomic level,” teams can make smarter calls earlier, rather than discovering incompatibilities after weeks of bench work.1
A recent work published in ChemPlusChem demonstrates that computational models can accurately capture how conditioning agents, peptide-based actives, and silicone alternatives bind to hair. This creates a rational foundation for optimizing performance rather than relying on heuristic blending.1
AI, Molecular simulation, and “Materials-intelligence.”
The materials-intelligence concept emerges at the intersection of high-throughput molecular simulation and AI models that learn from both computed and experimental data. MD simulations reveal time-resolved behavior of formulations at the nanometer scale, including micelle formation, phase separation, and the penetration of actives into lipid layers. Docking and quantum calculations capture specific binding and reactivity.
In recent cosmetics research, integrated workflows merge AI, molecular docking, and molecular dynamics to innovate bioinspired molecules, optimize hair-binding peptides, and identify environmentally friendly silicone alternatives for formulations.1
As Coquet notes, companies can “rapidly test thousands of potential formulations,” using simulation and AI to explore combinations that would be impractical to run physically. That doesn’t eliminate lab work; it rearranges it, so chemists spend more time validating the best options and less time manufacturing uncertainty.
On top of these physics-based engines, machine learning models absorb large datasets of ingredient descriptors, simulated properties, and lab results to predict stability, sensory attributes, and biological responses. Current research describes AI-driven predictive modeling across surfactants, polymers, preservatives, and other components, showing how models forecast rheology, texture, and shelf life while flagging potential adverse reactions such as allergic contact dermatitis.2,3
Here, Coquet says the goal is to “prioritize high-potential candidates” in ways that “reduce redundant experiments,” while also creating room to “identify novel cosmetic formulations that enhance performance and experiences.”
Accelerating Formulation while Cutting Waste
Image Credit: Cast Of Thousands/Shutterstock.com
Atomic-level simulations and AI accelerate development by allowing thousands of virtual experiments before any ingredients are weighed. High-throughput in silico screening ranks ingredient combinations by predicted miscibility, stability, and efficacy, so only the most promising formulations move into physical testing.
Companies like Nextmol use molecular modeling and AI platforms to characterize molecules, predict physicochemical properties, and automate the screening process for personal care applications, advancing only the top candidates to the lab.1,4
Coquet links this directly to sustainability.
It creates a sustainable workflow by streamlining development and providing a clear roadmap for testing, which in turn reduces time and resources spent on unpromising formulations.
Rudy Coquet VP Sales, Americas & Europe at Matlantis
There’s an elegance to this: Less wasted material, less wasted energy, and less wasted human attention.
Parallel developments in AI-driven formulation analytics help interpret historical data from previous projects. Machine learning models trained on legacy stability studies, sensory panels, and consumer feedback now suggest initial compositions, refine concentration ranges, and highlight potential incompatibilities early.
A recent report in the International Journal of Pharmaceutical Sciences states that such predictive models reduce trial-and-error, accelerate product development, and support data-backed design decisions that significantly cut development time, especially when embedded in integrated workflow tools used by global cosmetics companies.2,3
Coquet also argues that “high-speed computational platforms” can “strengthen collaboration,” enabling “continuous iteration” and “faster go-to-market timelines” – an important point, as faster decisions often come from shared visibility, not only faster computation.
Sustainability and Safety by Design
Materials intelligence changes how beauty brands approach sustainability by integrating environmental and safety considerations early in the design process, rather than at the testing stage. In silico toxicology and AI-driven safety prediction tools assess skin sensitization, irritation, and systemic exposure based on molecular structure and expected skin penetration. This reduces reliance on animal testing and aligns with regulatory trends favoring alternative methods.
Research in computational toxicology also suggests that AI models enhance reproducibility and transparency in hazard assessments.1,5 If simulation can steer formulation choices early, sustainability and safety stop being late-stage constraints and become design criteria from the start.
At the same time, AI and molecular simulation support the search for greener chemistries by comparing performance and environmental profiles of alternative ingredients. For example, Ferreira et al. describe how computational approaches guide the development of silicone alternatives through molecular modeling, mapping how structural variants change performance on hair while maintaining favorable safety and biodegradability profiles.
Additionally, industry reports highlight that AI systems can monitor regulatory changes, track ingredient bans across various jurisdictions, and propose compliant replacements in real time, making regulatory oversight a proactive element of R&D workflows.
Together, these tools support Coquet’s broader claim that the industry can “innovate more efficiently” without lowering standards as workflows can be engineered to filter, predict, and justify decisions earlier.1,2
Saving this article for later? Grab a PDF here.
Consumer Performance and Personalization
Atom-level simulations and AI also redefine performance and user experience in beauty products. Molecular simulations predict the interactions of active compounds with skin lipids, hair keratin, and pigment surfaces. This information forms the basis of the development of advanced delivery systems, encapsulation methods, and polymer networks that regulate release and film formation.
In cosmetics research, in silico analyses have been used to understand peptide binding to hair fibers and to screen bioinspired molecules with desired affinity and durability on biological substrates. These molecular insights translate into more targeted conditioning, color retention, or barrier support in final products.1
And because the models can evaluate options quickly, the same methods that improve performance can also help brands respond faster – Coquet notes that these advances help companies “respond quickly to changing trends and consumer demands.”
On the personalization side, AI models link molecular and formulation data with clinical, genomic, and imaging information to suggest ingredient combinations tailored to specific skin conditions or hair types. A recent Frontiers in Artificial Intelligence article on AI in aesthetic dermatology and cosmetogenomics shows how predictive modeling supports noninvasive treatment planning and product recommendations by integrating patient data with a mechanistic understanding of ingredients.
Combined with simulation-driven formulation design, this creates a feedback loop where real-world performance and diverse user profiles inform new generations of materials-intelligent products that deliver consistent, high-performing experiences to more consumers.6
What's in Store for Cosmetics R&D
This new era in beauty R&D is borrowing its best tools from pharmaceuticals and advanced materials and putting them to work on emulsions, polymers, pigments, and actives.
AI-driven molecular design systems like Chemistry42 show how generative models can propose new structures that satisfy multiple constraints at once, with computational prediction guiding the chemistry rather than trailing behind it.7
The next step is translation: Adapting these methods to cosmetic-specific datasets, performance targets, and regulatory realities, so “promising on paper” is successfully switched to “excellent in product.”
However, materials intelligence really only works when formulators, computational chemists, data scientists, and regulatory teams build and trust shared models. This means validating protocols, transparent assumptions, and ensuring simulations do more than decorate a slide deck, but decide what gets made.1
As high-speed platforms mature, the conjunction of digital and physical experiments will tighten; run virtual screens, test a focused set in the lab, feed results back, iterate again. In the long run this will lead to faster cycles and faster decisions.
References and Further Reading
- Ferreira, T. et al. (2025). Artificial Intelligence, Molecular Dynamics, and Beyond: Computational Insights In Cosmetics Research and Formulation Design. ChemPlusChem. DOI:10.1002/cplu.202500340, https://chemistry-europe.onlinelibrary.wiley.com/doi/abs/10.1002/cplu.202500340
- Speyer, R. et al. (2025). How AI is Transforming Cosmetic Formulation? EvalueServe. https://iprd.evalueserve.com/blog/how-ai-is-transforming-cosmetic-formulation-from-personalization-to-regulatory-compliance-and-sustainability/
- Punjab, S. (2025). Advancing Cosmetic Science Through Artificial Intelligence and Machine Learning: Personalization, Predictive R&D, and Sustainable Innovation. Int. J. of Pharm. Sci. Vol 3, Issue 11, 524-542. DOI:10.5281/zenodo.17522278, https://www.ijpsjournal.com/article/Advancing+Cosmetic+Science+Through+Artificial+Intelligence+and+Machine+Learning+Personalization+Predictive+RD+and+Sustainable+Innovation
- About Nextmol. Nextmol. https://www.nextmol.com/about-us/ [Accessed online: Jan 2025]
- Mazen, Al-Mohaya. et al. (2024). In silico approaches which are used in pharmacy. Journal of Applied Pharmaceutical Science, Vol. 14(04), pp 225-239. DOI:10.7324/JAPS.2024.154854, https://japsonline.com/admin/php/uploads/4213_pdf.pdf
- Haykal, D. et al. (2025). Cosmetogenomics unveiled: A systematic review of AI, genomics, and the future of personalized skincare. Frontiers in Artificial Intelligence, 8, 1660356. DOI:10.3389/frai.2025.1660356, https://www.frontiersin.org/journals/artificial-intelligence/articles/10.3389/frai.2025.1660356/full
- Ivanenkov, Y. A. et al. (2023). Chemistry42: An AI-Driven Platform for Molecular Design and Optimization. J. Chem. Inf. Model, 63, 3, 695–701. DOI:10.1021/acs.jcim.2c01191, https://pubs.acs.org/doi/10.1021/acs.jcim.2c01191
- Pathan, I. et al. (2025). Revolutionizing pharmacology: AI-powered approaches in molecular modeling and ADMET prediction. Medicine in Drug Discovery, 28, 100223. DOI:10.1016/j.medidd.2025.100223, https://www.sciencedirect.com/science/article/pii/S259009862500020X
Disclaimer: The views expressed here are those of the author expressed in their private capacity and do not necessarily represent the views of AZoM.com Limited T/A AZoNetwork the owner and operator of this website. This disclaimer forms part of the Terms and conditions of use of this website.