 |
| Getty Images; Rebecca Zisser |
Facial recognition technology is stepping into the medical arena, moving far beyond unlocking phones or tagging friends in photos. Now, AI-powered apps are being designed to detect pain in dementia patients, spot trauma in children, diagnose infections, and even predict biological age as a measure of overall wellness.
I recently tested Harvard’s new “FaceAge” app, which uses nothing more than photos to estimate a person’s biological age a quick proxy, its developers say, for health status. My results? Depending on the lighting and clarity of the photo, I looked anywhere between 27 and 38.
FaceAge is part of a wave of tools that turn selfies into diagnostic resources. Some apps can analyze facial cues to detect nasal congestion or seasonal allergies. Others watch for signs of drowsiness to keep drivers safe. There are programs that measure facial expressions for signs of illness, autism, or PTSD potentially reducing the need for children to recount traumatic experiences.
Since 2022, advances in artificial intelligence and faster, more powerful chips have supercharged the development of these tools. In 2025, face-scanning health technologies are becoming more sophisticated, promising earlier disease detection, more personalized treatments, and, according to some, the ability to forecast longevity.
“This is a medical biomarker, not just a novelty,” said Dr. Raymond Mak, a radiologist at Harvard Medical School and the lead developer of FaceAge, in an interview with Local press.
Still, experts are urging caution. “AI is moving into these spaces at high speed,” said Malihe Alikhani, a machine learning professor at Northeastern University. “The priority has to be ensuring these tools are safe, reliable, and truly beneficial.”
Your face can reveal far more than you think
 |
| Humans likely developed a third cone in our eyes to scan faces for signs of sickness or health. Now, AI developers hope to harness something similar. Anastasiia Havrysh/Getty Images |
For humans, reading faces as a measure of health isn’t new it’s instinctive. Research shows that about 30 million years ago, our ancestors evolved a third type of cone cell in the eye, enabling us to better detect subtle shifts in skin color.
“Rosy cheeks are often read as a sign of vitality, while greenish tones have long been linked to illness,” explained Brad Duchaine, a Dartmouth neuroscientist specializing in facial perception.
Science supports these associations. Flushed skin can signal healthy blood flow or high levels of carotenoids from a diet rich in fruits and vegetables. Conversely, stress, smoking, and environmental toxins can make people appear older by degrading collagen and accelerating wrinkles.
Plastic surgeon Dr. Bahman Guyuron studied identical twins with different habits, finding that the twin with higher stress and more exposure to toxins consistently looked years older. On the other end of the spectrum, some “superaging” centenarians appear decades younger than their chronological age, a sign of exceptionally healthy cells and organs.
Putting a face-scanning app to the test
 |
| FaceAge said I look 27.9 years old in this image - more than a decade younger than my actual age. Hilary Brueck |
Medical facial analysis tools are not brand new. The Face2Gene app, launched in 2014, helps clinicians diagnose genetic disorders and has been shown to outperform doctors in certain cases. PainChek, launched in 2017, measures micro-expressions to assess pain in nonverbal patients especially those with dementia and is now awaiting FDA clearance expected later in 2025.
 |
| FaceAge told me I have "healthy habits" because the computer thought I looked 10 years younger than my actual age. Mass General Brigham |
Since my writing often touches on aging, I decided to test Harvard’s FaceAge. While it’s still in the research phase, its ultimate aim is to use selfies for better diagnostics potentially tailoring cancer treatments to an individual’s biology or flagging hidden health risks.
FaceAge focuses on two areas: the nasolabial folds (between the nose and lips) and the temples (between the eyes and ears). These regions can reveal early signs of accelerated aging, which might indicate deeper health problems.
 |
| FaceAge said I look 38.2 years old in this image, close to my actual age. Hilary Brueck |
“If your facial age increases faster than your chronological age, it’s a serious warning sign,” Mak said.
I uploaded four photos. In a dark, grainy image, the app guessed I was 27.9 over a decade younger than my real age. A bright, makeup-free photo taken in harsh midday light made me look oldest. The other two one on a cloudy day, another in winter fell somewhere in between.
 |
| FaceAge: 33.1 years old Hilary Brueck |
This variability underscores a key limitation: lighting, sharpness, and image quality can dramatically affect results. As NIH neuroscientist Bevil Conway explained, removing fine detail from a face whether through blurriness, makeup, or lighting tricks tends to make it appear younger. Direct overhead sunlight, by contrast, emphasizes age lines.
 |
| Face Age: 36.8 Hilary Brueck |
So, am I biologically 10 years younger, or just one? The app can’t say for sure at least not yet.
Navigating the ethical gray areas
Even seemingly simple measurements, like estimating age, carry ethical concerns when AI enters the medical space.
“We’ve seen cases where companies roll out these systems for hospitals, border control, or law enforcement, only to realize later they were flawed or biased,” Alikhani warned.
The history of facial analysis is checkered. In 2017, a controversial Stanford project claimed to detect sexual orientation from photos a study widely criticized for conflating physical traits with environmental and cultural cues. Similarly, researchers at Shanghai Jiao Tong University once touted an algorithm that could identify criminals, raising alarms about pseudoscientific practices like physiognomy, historically used to justify prejudice.
Facial expressions are shaped by culture, gender, context, and individuality, making universal AI judgments risky. “Better healthcare means involving patients in decisions,” Alikhani said. “What happens when those decisions are outsourced to algorithms?”
The promise of face-based diagnostics is undeniable: quick, noninvasive assessments that could spot illness earlier than traditional methods. But until the science and the ethics are clearer, these apps should be seen as a tool in the medical toolbox, not the final word on your health.
