L’Oréal is bringing Maybelline’s virtual makeup try-on into ChatGPT through a collaboration with OpenAI. The experience combines conversation with ModiFace technology, allowing users to describe a look and preview products without leaving the assistant.
The partnership extends beyond one shopping feature into product discovery, advertising, life-sciences research and internal content creation. It shows how a consumer brand can treat a general AI interface as both storefront and operating platform.
Conversation gives virtual try-on a new entry point
ModiFace has powered augmented-reality beauty experiences since before L’Oréal acquired the company in 2018. L’Oréal reported more than 120 million uses of Beauty Tech services across 66 countries and 31 brands by the end of 2025.
Inside ChatGPT, a user can move from an imprecise goal—such as a natural evening look—to product exploration and visual preview. Conversation can explain preferences that filters and shade lists struggle to capture.
Accuracy matters because lighting, camera quality and skin tone influence the result. The experience should describe a preview as guidance rather than a guarantee that colour and texture will match reality.
Users also need control over facial images. Clear retention, consent and deletion rules are essential, particularly if a try-on image could reveal biometric or health-related information.
Product discovery and advertising meet at user intent
L’Oréal plans to improve how brands including Lancôme and Kérastase surface in ChatGPT in the United States. Several brands are also participating in OpenAI’s advertising pilot.
The proximity of recommendation and advertising requires transparency. Consumers should be able to distinguish a product selected for fit from one receiving paid placement. Otherwise a trusted assistant becomes an opaque sales channel.
Structured product data must include shade, finish, ingredients, price and availability. A conversational model should ground claims in current records and avoid inventing suitability for allergies or skin conditions.
L’Oréal’s e-commerce share exceeded 30% of sales in 2025, making AI discovery commercially significant. Retailers still need consistent service when the journey begins in a third-party interface, including stock, returns and human advice.
The same partnership reaches research and employees
L’Oréal is using OpenAI’s GPT-Rosalind model to study the skin microbiome, beginning with La Roche-Posay. The work seeks beneficial bacteria that may inform future skincare development.
Research use requires a different evidence standard from marketing. Model-generated hypotheses need experimental validation, reproducible records and expert interpretation. A fluent synthesis cannot substitute for laboratory proof.
OpenAI models will also support CreAItech, L’Oréal’s internal image and video platform. The company says 73,000 employees have received generative-AI training and already uses internal assistants.
Brand-history controls can help generated material remain recognisable, but rights, representation and approval still need human oversight. Faster production should not flood channels with repetitive or culturally unsuitable content.
Evaluation should include a broad range of complexions, ages, face shapes and lighting conditions. A preview that performs best for one customer group would turn convenience into exclusion. L’Oréal should publish accessible guidance about camera setup and provide an easy route to ordinary swatches or human advice when the virtual result is uncertain.
The collaboration connects consumer experience, scientific work and workforce tools under one strategic relationship. That creates leverage and vendor concentration at the same time, so L’Oréal should maintain portable data and evaluation.
Maybelline’s try-on may be the most visible feature because it makes AI tangible in a mirror-like moment. Its success will depend on restraint: realistic previews, clear commercial influence and respectful treatment of personal images. When those conditions hold, conversation can make beauty discovery more accessible without pretending that a model knows the customer better than the customer knows themselves.
