Can AI Build a Better Beauty Industry?

It’s in your phone, car, TV and social media. It’s there when you try to compose an email. You might be thankful for its help in your work life—or worried it will steal your job. And now, artificial intelligence is in your beauty products: not just in the obvious ways, like when you ask ChatGPT for skin-care tips, but in devices like mirrors and toothbrushes, the way new products are formulated, clinical trials, and the dermatologist’s office.

Industry experts are greeting these advances with a combination of excitement for the opportunities they’ll bring, skepticism over bold claims that science doesn’t support, and misgivings about where things could go next.

Asking an AI chatbot like ChatGPT for beauty tips—for example, to come up with product recommendations based on your skin type, if your primer will go with your foundation, or for an exact ingredients list—can be useful. “I think for these very specific questions where it can take a lot of time to dig through blog posts and scroll through social media, AI is a great tool to scrape what’s out there and give you a concise answer that maybe removes some of that trial and error,” says dermatologist Dr. Aegean Chan. “Or for a specific plan tailored to your lifestyle on, for example, how to use tretinoin, I think that is very useful to give people some guardrails. Doctors should be doing that—a lot of them just give the patient the prescription and tell them good luck.”

Could AI give ordinary folks a false sense of their own expertise and actually be dangerous? Yes. Chan uses the example of people ordering ingredients from overseas and compounding their own products, and she is horrified at one idea we’ve heard—performing home microneedling under the guidance of ChatGPT. “That is very concerning,” she says. “Medical-grade microneedling has the potential for serious infections and scarring. Also, I’m not sure it would be possible to purchase medical-grade equipment without a licence, at least from legit companies. The potential risks are real, and I’m not sure that AI will communicate that in a nonsycophantic manner. There is a reason that professional licensing exists.”

Things are changing rapidly, though, and she thinks her role might too. Accessing dermatologists’ services can be difficult and expensive, and there’s already a platform, Doctronic.ai, where you can get product recommendations for specific conditions. U.S.-based people can use the platform to connect with a board-certified dermatologist to create a treatment plan that may include prescriptions, though this is not available in Canada.

Diagnostic AI is also on the way. Derm by Skin Analytics is an AI medical device developed in the U.K. Available via the National Health Service, it helps people with suspected skin cancer get on the right treatment pathway. “People can take a picture of a bump, and it’ll sort of tell you, is this a basal cell carcinoma? Is this a squamous cell carcinoma, is this a melanoma?” Chan explains. Images are taken in a clinic and then sorted by the AI, with anything suspicious-looking being sent to a dermatologist for further checking.

Another similar idea is for reviews for people who have a lot of moles that need to be checked frequently—often every three months. “As AI gets more advanced, it will be able to monitor the pattern and look for evolution. Clinical examination is very subjective. So if the algorithm is trained well enough, I think that you can get a really high diagnostic accuracy that hopefully will reduce patient biopsies.” The main limitation for much of this AI-powered equipment is going to be cost, and Chan thinks it will be limited to big university hospitals and research centres, at least initially.

There is already some AI-powered equipment available to consumers, including the Swan mirror. This analyzes your face via questions and photos to track concerns like wrinkles, pigmentation, texture, oiliness, redness, acne, and UV spots, and then gives you a score between 0 and 100, followed by tips on how to improve it using products you can buy via the mirror. Chan isn’t keen. “I have concerns about body dysmorphia because it’s ranking your wrinkles, it’s ranking your pore sizes. Are those metrics that, without that mirror, you’re even clocking?” she asks. “It gets into that ‘lifemaxing,’ life optimization rabbit hole where it’s like, who is this for? Does this actually improve your health? Does it actually make your life better, or does it just give you more anxiety?”

The founder of the company behind the mirror, Colby Mitchell, says dermatologists manually rated thousands of faces to make the machine as smart as can be — but it’s unclear who those faces were. That’s something Rania Ibrahim, founder of SkinScience Analytics LLC, has thought about—not in the context of the Swan mirror, but when it comes to her specialist area, cosmetic testing.

“As far as AI being used in the testing world, it’s still pretty new, and in my opinion there’s a lot of hand waving when it comes to it,” says Rania Ibrahim, founder of SkinScience Analytics LLC. “The way that they’re incorporating AI right now is with software programs where you would take an image of an individual and then they would determine the age of that individual based on these algorithms that they’ve created. And someone would use a skin-care product for, say, eight weeks, and then they would take another image and do the same thing.”

The goal is to allow beauty brands to claim their products have reduced concerns like wrinkles, acne or hyperpigmentation, for example. But Ibrahim isn’t convinced.

___

“My hesitation with this software analysis or AI analysis is we have no idea what kind of data sets they’re using. We don’t know how many people, we don’t know how diverse the population is, because diversity, skin types, environmental conditions, all of these things play a factor in how someone ages and how you’re determining the age of an individual.” —Rania Ibrahim, founder of SkinScience Analytics LLC

 

One of the most exciting areas that scientists are seeing AI used in is in product formulation. Boris Zion is a software engineer and CEO of Cosmetic Makers, a contract manufacturer of beauty and personal care products, so he was perfectly positioned to create an AI-powered formulation platform. He founded CM Studio+ two years ago to help professional formulators and students develop beauty products. You write a detailed description of the product you need—for example, a light-textured serum that targets hyperpigmentation, has a lavender scent, and complies with EU cosmetic regulations. Within a few minutes, you’ll get a formula complete with percentages of ingredients, based on your specification. You can interact with the AI, which uses proprietary training data as well as Gemini and ChatGPT, to tweak the formula, source ingredients, figure out costs, packaging, and more. “I could ask ChatGPT to do this for me, and it would come up with an answer, but this has context into inventory availability, cost per kilogram for different ingredients, effectiveness, overall performance, stability and feedback from consumers,” Zion explains. You can also reverse engineer a product formulation: take the ingredients list for a product and get a structured list of how it’s made.

You don’t have to be a professional to use the platform, but you do need expertise to actually figure out if a formula makes sense and make samples, let alone manufacture them at scale. Jen Novakovich, a cosmetic scientist and science communicator has seen a demo of it herself and is impressed. “The formula is something that when I look at it, it would be a good starting point to develop off of and then iterate off of,” she says. “This would really expedite the process.”

She’s seeing AI creeping into many corners of the formulation world, including in developing completely novel ingredients. One example she cites is in fragrance, where master perfumer Christophe Laudamiel and a team at AI startup Osmo has synthesized a single molecule that smells like fresh cantaloupe. Traditionally, melon notes require many molecules, expensive ingredients, and can be skin-sensitizing and unstable. “Now there’s a melon aroma chemical that is going to be widely accessible and easy to formulate with. So definitely there’s opportunity for ingredient innovation.” Predictive formulation, in which AI computes how molecules are likely to behave before a scientist has even entered a lab, is the next big innovation.

Novakovich’s focus these days is more on scientific communication than formulation, and there she’s seen AI content proliferate in concerning ways. “It’s creating a real issue in independent, good science communication,” she says, alluding to how creators can produce vast amounts of content that looks solid and scientific but is sometimes total nonsense. “It’s flooding social media and creating more drain on the people who are doing independent science communication.” She says people on social media are finding ways to weed out anyone using AI. “Right now, people are looking for imperfection. They’re looking for human language. I’m a master of typos, and people now like my content more because I have typos everywhere — it’s proof that I’m just a human creating my content.”

SHARE