AI and Beauty: A Double-Edged Sword
Personalization, Empowerment, and the Risk of Unrealistic Standards
Disclaimer: I built StyleSense.io, a platform built to personalize beauty recommendations. I’ve thought a lot about how to productize AI for beauty use cases, and have tried to do so in a thoughtful way that avoids the risks I’ve outlined below. But yes, I explicitly have a vested interest in this space. (Read more about my founding story here).
I don’t know where I stand yet on this issue. I’m both excited by AI’s potential for personalized beauty, and mindful of the risks that come with a world of digital images and filters.
Let’s step through the pros and cons:
The Positives: More Intentional Consumption
- Personalized Beauty Recommendations: AI has the power to help cut through the noise of beauty tutorials and constant advertising. Rather than mindlessly pushing products or looks that may not suit you, AI can curate options tailored to your unique features — whether that’s your skin tone, face shape, or personal preferences. Look to the popularity of AI color analysis or customized skin care as an example — by narrowing choices, AI can create a more intentional and empowered beauty consumer at scale. One who chooses products or content because they are personally relevant, rather than because of an ad or influencer’s recommendation.
- Virtual Try-Ons for Experimentation: Another exciting possibility is the ability to experiment with looks virtually. This allows people to play with styles or products without having to commit to purchasing them. Whether it’s trying on lipstick, eyeshadow, or even a full outfit, these AI tools make it easier for consumers to express and explore without buying something they may not love. Virtual try-ons also offer a lower-stakes environment for people to find what works best for them without the pressure of trends or conforming to global beauty standards.
The Negatives: Reinforced Bias and Unrealistic Standards
- Reinforcing Beauty Biases: At StyleSense, I use AI to classify face shapes, skin tones, undertones and eye shapes. For these quantified classifications, it is feasible to test performance and accuracy across diverse groups — however, there are significant challenges in evaluating generative tasks like virtual try-on, since there is no easily defined “correct” result to benchmark the results against. (I actually set out to build try-on features in stylesense.io but abandoned it due to this and many other reasons, which I may write about at a later point). AI, like any technology, is built on data — and if that data is skewed, it can reinforce existing biases, which is particularly salient when we talk about it using it for different skin types and tones.
- Accelerating Unrealistic Beauty Standards: While unrealistic beauty standards have always been present (eg. magazines), the advent of AI-based filters like TikTok’s Bold Glamour gives me pause. Unlike a magazine cover, which you can set aside and forget, an AI-powered filter can subtly alter your appearance in real time, often without you even realizing it. As these tools become more accessible, younger generations, in particular, may struggle to distinguish between digitally altered beauty and reality. It’s something I experienced firsthand, with the popularization of snapchat filters. I wonder how our self-perception will change as this technology becomes more and more undetectable and we accept these idealized edits as reality. I’m sure the plastic surgery industrial complex is very happy.
A Path Forward: Balancing Innovation with Responsibility
To create a future where AI benefits beauty consumers, it’s crucial to:
- Develop more inclusive AI models that account for a wide range of features, skin tones, and identities.
- Educate consumers about the reality behind AI filters and virtual try-ons, helping them navigate digital beauty spaces with a critical eye.
- Foster a beauty culture that celebrates individuality, diversity, and authenticity over unattainable perfection.
I do believe that AI can be a tool for empowerment —obviously, as I’ve built a website pushing that forward.
But I also want to wait and see, and I hope that this space develops with thoughtfulness at its center.