I have a huge pimple on my cheek when I try Olay’s skin analysis app, Skin Advisor. The app directs me to take a selfie and answer a few questions, like what worries me most about my skin — wrinkles or acne, for example — and what kinds of products I use on a daily basis.

I’m prepared for the obvious news: I’m having a breakout. But within a few seconds, the app alerts me to an issue I didn’t know I had. According to Skin Advisor, my problem area is around my mouth, likely because of my smile lines, and I’m directed to buy a $22 anti-aging cream.

This is curious, since TroveSkin, the Indonesia-based artificial intelligence app I tried before this one, analyzed my selfie and concluded that my biggest skin problem was texture, and that I should be using exfoliating, anti-aging, and brightening products.

Skin Advisor and TroveSkin are two of about a dozen apps from startups and veteran skin care companies that use artificial intelligence (A.I.) to spot your skin woes, provide personalized product recommendations, and track your skin over time. In 2018, Neutrogena released a device and app package called Skin360, which uses a magnifying lens attached to your smartphone to deliver product suggestions. Startup Proven recently launched a personalized skin care line that draws from medical journals and reviews to suggest products based on your responses to an online assessment.

The global cosmetics market is estimated to grow to $805 billion by 2023, with skin care accounting for the largest share. Between drugstores, online retailers, and beauty stores, it’s hard to know what you really need or want. Do I have blackheads or redness? Should I try to brighten my skin or start an anti-aging regimen? How do I make my skin look “dewy,” whatever that means? Hyperpersonalization using A.I.-driven technology has caught on in consumer products (think Google Home, Netflix, and smartphone facial recognition), so it makes sense that businesses see customizing your skin care routine as a logical application.

Diana de la Iglesia Jiménez, an A.I. and computer engineering expert at the Bioinformatics Unit of the Spanish National Cancer Research Centre, a public research center, says the technology used in skin care apps is the same machine-learning technology used in other industries. “They collect data about you and predict your current needs or the potential features of your ‘ideal’ product,” she says, adding that she understands the technology’s appeal. “If these products are capable of analyzing your specific needs and features, they could bring you a solution more adapted to your real needs.”

TroveSkin, for example, encourages you to take a daily selfie so you can follow how your skin is changing over time. I could log in throughout the day to let the app know how I was feeling, how long I slept, how much I exercised, and what I ate throughout the day.

Some apps also guess your age depending on your skin’s appearance—what’s known as your “skin age”—with varying degrees of accuracy. TroveSkin, the first app I tried, guessed my age was 23, underestimating it by six years. When I used my laptop camera, Olay’s Skin Advisor correctly guessed I was 29. But when I tested it again using the app on my smartphone, it guessed I was 35.

Personalization is not just a trend. It’s going to be the future of all consumer products.”

Proven co-founders Ming Zhao and Amy Yuan developed their business in response to their own skin troubles. Before starting Proven, Zhao was working 16-hour days at a private equity job in Hong Kong and “using all kinds of different miracle-promising skin care products, and I felt like nothing made a dent,” she says. “Eventually, someone recommended I see a skin guru to make personalized skin care for me. That was the only time I saw results.”

When Zhao met Yuan, a computational physicist, at Stanford University, the duo started building a deep-learning skin care database using scientific studies and consumer testimonials. It now includes more than 8 million consumer reviews, 100,000 skin care products, 20,000 ingredients, and 4,000 scientific journals.

Here’s how it works: Customers answer a list of questions about their skin concerns, how their skin reacts to products, and even what they eat. After you complete the quiz, Proven’s machine-learning algorithms scan the database to spit out information about your living environment and lifestyle and how they affect your skin and to match you with ingredients. Zhao says the company partnered with a dermatologist to work on the questionnaire and create the formulations it sells.

When I tried Proven, I was told I’m not drinking enough water, which is leading to dehydration. Since I live in Dallas-Fort Worth, which has “unhealthy air quality,” Proven told me I’m exposed to pollution particles that cause wrinkles and age spots.

To address these environmental factors along with my acne-prone, oily skin and hyperpigmentation, Proven suggested a customized three-product routine that includes ingredients like sulfur (to deal with acne), hyaluronic acid (to retain moisture), and mandelic acid (to exfoliate the skin for a smoother, clearer appearance). The three products cost $195 for a two-month supply.

“Personalization is not just a trend. It’s going to be the future of all consumer products,” Zhao tells me. “With Proven, our products are tailored to you [based on] genetics, skin concerns, and environmental factors.”

I ask dermatologist Fayne Frey if personalization apps are the future of skin care. Her response is a resounding no. “Apps have little if any clinical skill other than programmed, one-size-fits-all algorithms,” she says.

Frey takes issue with some of the questions asked by Proven. Most consumers probably don’t know what’s wrong with their skin or how to answer questions about their level of stress, which is subjective, she says.

Frey’s other issue with Proven, Skin360, and similar apps is that they lead consumers to purchase the company’s own products. The tech, according to Frey, is more about marketing than about giving people science-backed recommendations — the latter of which could actually be helpful. It’s a “market that’s feeding on people’s insecurity,” she says.

Frey is attempting to create a nonbiased database herself. In 2014, she launched the website and product selector FryFace. Based on a few questions, consumers can find skin care options. She doesn’t make any money from the site and adds products manually. In her opinion, there are plenty of products already on the market that work just fine.

Dermatologist Dee Anna Glaser, of St. Louis University, says that if A.I. apps get people thinking about taking better care of their skin, that’s not a bad thing. Unfortunately, A.I.-backed skin care apps are not cheap. For $195, Proven’s customized kit includes one cleanser, one sunscreen, and one night cream. It costs $50 to get Neutrogena’s Skin360 app and the magnifying lens required to scan your skin. Olay’s Skin Advisor appears to be one of the few free-to-use A.I.-driven apps currently available.

“All of these A.I. devices lead you to buy some kind of skin care. And if it’s working, that’s great,” Glaser says. “But what I see at my practice is that patients spend hundreds of dollars [on products that don’t work], and then they’ve delayed the opportunity to get better. People with true skin problems would be better served more quickly, accurately, and less expensively” if they went to see a dermatologist.

A.I. expert de la Iglesia says there’s another issue: If these apps collect low-quality data (an image that’s too blurry, a question that wasn’t answered properly, or product reviews that are fake), a machine-learning algorithm does not work properly. “I think that the collected data could be of low quality, and so [could] the predictions,” de la Iglesia says.

A.I. is making an impact, however, on the detection of skin diseases. Glaser says there are already devices in dermatology offices that use A.I. technology to help doctors detect skin cancer and determine whether a biopsy is needed. As the technology becomes more accurate, it might even be applicable for consumer use.

While these apps may provide a good starting point for people who are new to skin care, I wouldn’t use them again unless I was looking to buy a product from the brands that offer them (say, I needed help choosing between different Olay moisturizers). The aim to help consumers select products that are more likely to fix their skin problems is a lofty goal, but in my many years of experimenting, I’ve found which ingredients work for me and which ones don’t — and I’m not sure that an app can be better at selecting products than I am.