Divyanshu Srivastava, Shikha Baghel Chauhan, Rajan Swami, Indu Singh
AI and ML are reshaping cosmetic formulation from an empirical, trial-and-error process into a data-driven, predictive discipline. While challenges related to data quality, algorithmic bias, regulatory harmonisation, and validation remain, the trajectory indicates that AI will become an indispensable partner in designing safer, more effective, and more personalised cosmetic products.
INTRODUCTION: Cosmetics and cosmeceuticals serve both aesthetic and hygienic functions, yet traditional formulation approaches remain resource-intensive, empirical in nature, and limited in their capacity for personalisation. The integration of artificial intelligence (AI) and machine learning (ML) into cosmetic science offers a transformative alternative, enabling faster ingredient screening, predictive safety and efficacy assessment, and individualised product design.
METHODS: A systematic literature search was conducted across PubMed/MEDLINE, Google Scholar, ScienceDirect, IEEE Xplore, ResearchGate, and Google Patents, covering publications from 2013 to March 2025. Search terms spanned AI-driven formulation design, predictive modelling, ingredient optimisation, toxicity prediction, personalised skincare, and regulatory considerations. Peer-reviewed articles, technical studies, and patents meeting pre-defined inclusion criteria were selected.
RESULTS: This review systematically examines AI and ML applications across the cosmetic formulation pipeline from ingredient selection and stability prediction to toxicity screening, safety assessment, and personalised recommendation systems.
DISCUSSION: Unlike previous reviews that primarily address diagnostic or imaging applications, this work's particular emphasis is on formulation science: predictive modelling of preservative efficacy, in silico toxicity assessment using quantitative structure-activity relationship (QSAR) and read-across tools, AI-assisted skin analysis, and the algorithmic frameworks underpinning these advances. Comparative discussion of convolutional neural networks (CNN), artificial neural networks (ANN), support vector machines (SVM), random forests, and gradient boosting approaches contextualises their strengths and limitations within cosmetic applications.
CONCLUSION: AI and ML are reshaping cosmetic formulation from an empirical, trial-and-error process into a data-driven, predictive discipline. While challenges related to data quality, algorithmic bias, regulatory harmonisation, and validation remain, the trajectory indicates that AI will become an indispensable partner in designing safer, more effective, and more personalised cosmetic products.