You probably know those days when nothing feels quite right. You keep opening the same apps, reply to messages in bursts and keep returning to your phone. Most of us chalk it up to a mood we can't quite name.
A new AI tool picks up those tiny shifts in phone habits and reads them as signs of stress. Researchers at KAIST in South Korea built a system called SHIRBT-MMF that spots stressful days from how a person uses their phone, with no watch or ring attached.
The study appeared in the journal Proceedings of the ACM on Interactive, Mobile, Wearable and Ubiquitous Technologies in December 2025.
Rather than counting minutes spent in each app, the system looks at what the person is actually doing. Think messaging a friend, browsing the web, paying for something or writing an email.
Also read: A new type of Alzheimer's drug shrank plaques in mice without a brain-swelling signal
The team identified 29 types of repeated behaviors, and most participants had a routine of around seven. During stressful periods, the fingerprint of those behaviors shifted in ways the users themselves did not consciously register.
How accurately phone habits flagged a stressful day
The personalized AI hit 75.9 percent accuracy for participants tracked around a month, and 73.8 percent for those tracked around 105 days, according to KAIST. Both numbers beat conventional app-usage analysis by several percentage points.
The study was small. It followed 26 university students and office workers in their 20s and 30s, so the result is a proof of concept, not a finished product.
Just one behavior category, messaging friends on chat apps, outperformed the traditional app-usage approach on its own. The pace, length and timing of messages carry more stress signal than most people would expect.
Also read: Is running bad for your knees? Here is what the evidence actually says
What this could mean for passive mental health monitoring
The researchers argue that a tool like this could reach people who never actively seek support. A phone that sits in a pocket all day could flag a rough stretch and nudge the user toward sleep, a break or a check-in, before stress becomes something louder.
"We viewed the smartphone not simply as a device for measuring usage time," Professor Uichin Lee of KAIST said. He described it as a space where everyday communication, information seeking and financial activities are repeatedly carried out.
Privacy was built into the design. The system did not store any text or message content, only behavioral metadata such as how often someone typed and how screens moved. Chat partners and contacts were hashed, so no person could be directly identified from the data.
Stress monitoring on a phone is still research, not something a reader can download today. The team now plans to test the system across more varied age groups and more types of smartphones before it could be built into any app or clinic tool.
Also read: 8 warning signs of perimenopause most women don't expect in their 40s
Also read: Berries, poultry and 2.5 extra years: What the MIND diet may do to your brain
This article is made and published by Jesper Bengtson, who may have used AI in the preparation.