When developing a cosmetic product, product developers consider which ingredients to use and in what amounts. The resulting ingredient composition and proportions make up the product’s formula. Once a sample has been prepared, characteristics such as viscosity—which describes how thin or thick the product is—and spreadability are evaluated, and the formula is adjusted as needed.
Recently, some global cosmetics companies and ingredient suppliers have begun using AI in these development processes. AI can help analyze formulation data accumulated over time and evaluate ingredient combinations that may deliver a desired sensory feel.
How, then, can AI assist in evaluating ingredients and formulas? In this article, we examine examples of how AI is being used in product development, as well as what needs to be checked as a developed formula moves into sample preparation and commercial production.
1. How Are Global Cosmetics Companies Using AI?
In March 2026, L’Oréal announced an expansion of its AI collaboration with NVIDIA. The collaboration involves using computational modeling to predict ingredient properties and interactions, supporting research into the development of new formulations. The announcement identified photoprotection and skin tone management as key areas of application.
Earlier, in 2025, L’Oréal announced plans to work with IBM to develop an AI model using formulation and ingredient data. The goal is to support researchers in product development tasks such as identifying suitable formulas for new products and adjusting existing formulas.
These developments illustrate how the cosmetics industry is increasingly exploring the use of AI in product development for tasks such as predicting ingredient properties and evaluating formulas.
2. AI Can Predict Formulation Properties Using Development Data
AI can be trained on data about ingredient combinations, ingredient ratios, and experimental results to predict the properties of formulations with different compositions.
In a study published in ACS Omega in 2026, researchers prepared and tested 27 cosmetic formulations made with water and oil. They varied the proportions of ingredients including sunflower seed oil and shea butter, as well as the proportions of the viscosity modifiers carbomer and xanthan gum.
Twenty-four hours after the samples were prepared, the researchers used instruments to measure the viscosity and flow behavior of each sample.
They used the resulting formulation data and experimental measurements to train an AI model. The model was then used to predict the properties of additional formulas that were not included in the training data, and the predictions were compared with actual measurements.
The AI’s viscosity predictions for these additional formulas were close to the measured values. However, the study also identified limitations, as the model did not fully predict variations in some evaluation parameters.
The study demonstrates that AI can be used to predict some formulation properties within a defined set of ingredients and composition ranges. Whether comparable accuracy can be achieved with other ingredients or product types would need to be evaluated separately.
When organizing development data for AI training, it is useful to record not only the ingredients and their proportions but also the experimental conditions and measurement methods.
3. AI Is Also Being Used to Evaluate Ingredient Combinations
Ingredient supplier BASF offers Emollient Maestro, an AI-powered service that helps evaluate combinations of emollients. Emollients are ingredients used to soften the skin and adjust the sensory feel of a product.
BASF explains that the service helps users identify alternative ingredient combinations that can provide properties and sensory characteristics similar to those of existing ingredients. For example, when replacing certain silicone ingredients, users can compare how different emollients might be combined.
In October 2026, BASF announced that the service would also provide information on how readily ingredients evaporate. The company explained that this information can be used to assess how a product’s sensory feel changes over time after application.
When using tools like this, it is helpful for brands to define which aspects of the product’s sensory feel they want to maintain or change. They can then tell the manufacturer whether they want the product to retain a smooth, easy-to-spread feel after an ingredient is changed or whether they want to adjust the feel left on the skin after use.
4. AI Predictions Must Be Confirmed with Actual Samples
For example, suppose a brand uses AI to predict the viscosity of several formulas while developing a lotion. It may then select the formula predicted to be closest to the target viscosity and prepare a sample.
After the sample is prepared, its viscosity should be measured and compared with the AI prediction. The sample should also be evaluated for spreadability and stickiness. A product may fall within the target viscosity range without necessarily providing the desired sensory feel.
Depending on the product’s characteristics, it may also be evaluated for phase separation during storage and changes in color or odor.
Predictions of individual properties, such as viscosity, are not sufficient to determine the overall quality or safety of the finished product. Appropriate tests and evaluations must therefore be conducted using actual samples.
5. What Should Be Checked When Moving to Commercial Production?
Even when AI has been used to evaluate a formula, the equipment and manufacturing conditions required for commercial production must also be considered. Commercial production may involve different equipment and manufacturing conditions from those used to prepare small samples in a laboratory.
Factors such as the order in which ingredients are added, mixing time, mixing speed, and temperature should be reviewed. The resulting product must then be checked to confirm that it meets quality specifications, including the target viscosity and appearance.
When discussing development with their manufacturer, brands should also clarify what verification steps and preparation time will be required between finalizing the sample and beginning commercial production.
Conclusion
AI can be used to analyze accumulated formulation data and evaluate new ingredient combinations or formulas. Turning this analysis into an actual product still requires researchers’ judgment, sample evaluation, and quality control throughout production.
Neo Mirae works with brands throughout the process—from formulation development to production—based on the sensory feel and product characteristics they want to achieve. If you have a product in mind, work with Neo Mirae to define a clear development direction.
Start your project with us.
