If skincare is a science, it will sooner or later be reinvented by AI | Early-stage projects

Author | Zhong Yixuan

Imagine a scenario - without going to an institution or even leaving home, anytime and anywhere, just facing the phone camera, you can obtain a professional and traceable skin diagnosis report within just a few seconds, and enjoy personalized skincare recommendations and consultation services.

This scenario has already become a reality today. "You Look So Good Today" is such an app that 36Kr recently got in touch with. Wu Liang, the founder of "You Look So Good Today", majored in computational biology and was an influential creator in the biology field on Zhihu in the early days. Wu Liang told 36Kr that his initial inspiration for starting a business stemmed from sharing skincare knowledge on Zhihu. After gaining a large number of users' attention, he keenly observed a serious mismatch between supply and demand in the skincare market - there is a large gap between people's strong demand for personalized skincare and the high scientific threshold of skincare itself.

According to Mintel's research data, the skincare needs of global users are accelerating the transformation towards refinement and personalization. 62% of consumers hope to customize personalized skincare products according to their own skin conditions. However, as a dynamic organ, the skin is closely related to the immediate state and the environment. Relying solely on standardized products or experience - based recommendations from bloggers cannot meet personalized needs.

Wu Liang believes that although the skincare trend in China has changed many times, the essence of skincare remains the same, which is to provide a personalized closed - loop from "precise detection" to "customized service", and this has always been the most core pain point for every user.

"You Look So Good Today" app aims to provide users with a complete set of AI skin intervention solutions from "detection - discovery - consultation - solution". "Our positioning is an online dermatology and cosmetology hospital," Wu Liang told 36Kr.

Therefore, ensuring the accuracy of skin image collection, detection, and analysis is the core competitive ability of the team. Wu Liang introduced that over the years, the team has continuously refined the skin - testing model and maintained an average iteration speed of once every 1 - 2 months. Currently, the model has accumulated more than 200 million skin data, and both the data scale and algorithm ability rank first globally.

Display of the skin - testing model

Display of the skin - testing model

Specifically, the above - mentioned skin - testing model covers the entire skin management cycle through high - precision collection, professional - level image processing, and an algorithm model of "professional evaluation of product ingredients + matching degree". Taking the collection algorithm as an example - for a mobile phone image, factors such as light, posture, lens parameters, and even the phone model can affect the final image. To ensure accuracy, the team has tested all mobile phone models on the market.

"This is much more difficult than expected. When we first launched, we only had the iOS version. Later, when we expanded to the Android version, we found that there were too many models. So we specially developed an algorithm for image correction and pre - trained it based on 26.54 million labeled images. Currently, we have achieved a missed - detection rate of less than 1%, an accuracy rate of 93.6% for acne grading, and an accuracy rate of 95.7% for pore classification, comparable to professional - level equipment," Wu Liang said.

In real - time image processing, "You Look So Good Today" can process 12 million images per second, while the general processing time in the industry is 1 - 2 minutes, showing a significant efficiency advantage.

Wu Liang also mentioned that the result analysis of AI skin testing needs to be closely coordinated with dermatological medicine, which requires the team to transform the research results of skin science mechanisms into precise skin - testing algorithms. Currently, the detection results of "You Look So Good Today" are highly consistent with those of medical equipment, and the team has published several papers in top dermatology journals. For example, they cooperated with Peking Union Medical College to study the formation factors of dark circles under the eyes of millions of Chinese users and the differences between genders.

Articles published in top journals in the skin field

To further improve the accuracy of detection results, "You Look So Good Today" also collects information other than skin images from users to simulate the process of a face - to - face consultation as much as possible, such as sleep and diet, whether the user just got up, is about to go to bed, is in the recovery period after medical aesthetics, or is in a sensitive period. "Just like a dermatologist makes a judgment through 'observation, auscultation, inquiry, and palpation', we also ask users to fill out a questionnaire before the high - definition camera detection to find out other factors that may affect skin detection as much as possible," Wu Liang said.

In addition to precise detection, quantitative tracking is another major differentiating advantage of the product. In the past, whether in hospitals or medical aesthetics institutions, the skin reports or data detected by users were generated only once and could not be retained freely. "You Look So Good Today" can track the long - term trend changes of users' skin. The long - term retention rate of users who have undergone medical aesthetics on the app has exceeded 45%.

Currently, the team also provides users with customized skincare solutions, that is, combined with AI corpus, it offers one - on - one consultation services, which have been introduced into the WeChat ecosystem. Although everyone's skin is different, the skincare industry is ultimately a scientific field, which is the core reason why the AI corpus can assist in consultations. In addition, introducing the service into WeChat not only means low - cost and high - frequency user reach but also helps to further accumulate corpus and improve the algorithm.

36Kr learned that the total number of users of the "You Look So Good Today" app has exceeded 18 million, of which 14 million are from organic traffic, coming through word - of - mouth. Moreover, 10 million users have actively participated in the evaluation of cosmetics in the app. The 6 - month retention rate of the app reaches about 30%, indicating strong user stickiness.

In addition, based on the existing software, the team is currently extending to hardware devices.

Wu Liang mentioned that the founding team is currently developing a smart makeup mirror, which will integrate the detection function and service system. Now it is in the prototype stage, and the expected selling price is 800 - 1000 yuan in China and 150 US dollars overseas. Currently, the home - use detection smart mirror developed by the team has two modes: skincare and makeup. The former can provide one - click AI skin testing, recommend customized skincare steps and products, and predict future skin problems; the latter provides AI makeup guidance and simulates lighting effects in different scenarios.

But the question is, why would users pay for a smart mirror worth about a thousand yuan when they already have the app and can even get free detection and consultation?

Wu Liang revealed that the plan for the smart mirror product germinated several years ago, and the team finally made up its mind in 2025 due to breakthroughs in two major capabilities.

One is the improvement of sensor capabilities. Compared with mobile phones, the sensor capabilities of the smart mirror are more accurate. It not only supports visible light detection but also invisible light detection. For example, it can use ultraviolet light to detect information about bacteria and skin pigmentation on the skin surface. It can even predict acne problems in advance by monitoring fluctuations in skin indicators.

The other is the breakthrough in AI technology, which enables more convenient interaction on the terminal. "No one wants to leave fingerprints on their mirror, but AI technology allows people to interact through voice, and the mirror can present or play skincare suggestions. If the user didn't sleep well last night and has dark circles under the eyes, the smart mirror will suggest applying the most suitable eye cream in the morning and doing appropriate massage to reduce swelling; or provide acne prevention plans; or recommend the most suitable mask according to the current skin condition; and recommend the most suitable foundation according to the user's skin tone and skin type," Wu Liang said.

In short, the smart mirror is a physical carrier of the app's algorithm and model capabilities. It seizes the high - frequency scenario of users spending 20 minutes in front of the mirror in the morning and evening, broadens the product's usage scope, and more conveniently obtains more data to improve the algorithm capabilities.

In response to users' skin data and problems, the team has also launched its own brand, Lakelab, with a total of 15 SKUs, focusing on efficacy - based products. The current sales have exceeded 30 million yuan, and the annual repurchase rate is as high as 63%. Wu Liang mentioned that the above sales conversion only relied on 1% of the app's recommendation traffic without any additional advertising expenses. The team relied on the advantage of the skin data accumulated by the app in the past to form accurate and high - quality product matching, resulting in efficient conversion and remarkable repurchase data.

Display of the skincare product matrix

36Kr also learned that the "You Look So Good Today" app went global in August 2025. Wu Liang mentioned that the biggest challenge in going global is that the skin conditions and skin tones of overseas users are different from those of Chinese people, so the algorithm needs continuous improvement. However, the subscription - payment conversion rate of the overseas version has reached 8% - 10%. "Through the overseas launch, we unexpectedly found that users' willingness to subscribe is not bad, which also gives us some confidence in launching paid subscriptions in China," Wu Liang said.

Regarding the founding team, the team of "You Look So Good Today" has about twenty members and has been relatively stable. Most of them have accompanied Wu Liang since the start of the business. When it comes to the goals for the next 3 - 5 years, Wu Liang said that he hopes to help users effectively solve their skin problems. "In the future, 'You Look So Good Today' will not just be a detection tool, but a Doraemon, like a good friend who understands your skin best and can give you the most suitable skincare plan and makeup guidance. That's why the logo of our app is a kitten instead of a camera."

For this reason, he also wrote in his personal signature on his corporate WeChat, "Help solve the skin problems of half of the acne - prone and sensitive - skin people in China." So far, these are two problems recognized by the skin science community and the industry as having the widest affected population, the most complex causes, and being the most difficult to completely eradicate. It is also the ultimate mission of the team to promote the skincare health of the general public through AI technology.