At some point in the coming years, someone may receive an unexpected message on their phone. They would not have asked for help. They would not have an appointment scheduled. They might not even think that anything is wrong.
Yet the system would have detected a gradual transformation. Nothing dramatic would have happened overnight. Their life would simply begin to change. They would leave home less often, their daily routines would become less regular, and activities that once formed part of everyday life would slowly disappear. From the inside, each change would seem insignificant. Viewed through months of accumulated data, however, it might look like the beginning of something more important.
The scenario sounds futuristic, but it reflects the direction that part of mental health research is already taking.
A 2025 review analyzed 112 studies on digital phenotyping, the use of smartphones and connected devices to monitor human behavior. The review found that mobility, physical activity, sleep, and social interaction were among the most commonly studied variables. At the same time, researchers identified substantial variation in how these data were collected and analyzed, making comparisons across studies difficult [1].
The question running through much of this literature is straightforward: can these data detect a relapse before either the patient or the clinician recognizes it?
Psychiatric relapses rarely emerge suddenly. In many cases, they develop over weeks or months through small changes in daily routines. Sleep becomes irregular. Physical activity declines. Movement patterns become more restricted. Social interactions begin to diminish. Individually, none of these signals may be sufficient to justify a clinical intervention. Taken together, however, researchers are investigating whether they can serve as early indicators of deterioration.
A systematic review published in 2025 examined studies that used smartphone and wearable data to track patients with depression over time. Researchers found consistent associations between worsening depressive symptoms and reduced geographic mobility, lower levels of physical activity, and disruptions in sleep patterns. Several studies also suggested that these variables could help anticipate future changes in symptom severity [2].
Another review published the same year identified the most commonly used indicators in digital phenotyping research. These included accelerometer-based measures of physical activity, daily step counts, heart rate, sleep duration, and sleep quality [3]. Remarkably, much of this work does not rely on molecular biomarkers or laboratory tests. Instead, it draws on routine behaviors that people generate passively as they go about their daily lives.
Researchers are increasingly treating these behaviors as behavioral biomarkers. Just as changes in certain blood parameters may signal a physiological process, sustained reductions in mobility or persistent disruptions in sleep could reflect changes in mental state.
In 2025, an analysis published in the Proceedings of the National Academy of Sciences reviewed developments in the field and highlighted how smartphones can capture objective measures of human behavior outside clinical settings. The authors argued that these tools could complement information gathered during medical consultations and help identify periods of increased vulnerability [4].
But the most important question remains the same: do these systems actually work when the goal is relapse detection?
A 2026 review evaluated digital technologies specifically designed to identify relapses in mental health conditions. The studies examined smartphones, wearables, and remote monitoring systems aimed at detecting early clinical deterioration. The authors found promising results, particularly in conditions where behavioral changes precede more visible symptoms. However, they also noted that most studies remain relatively small, heterogeneous, and difficult to compare, limiting the strength of current conclusions [5].
The next stage of research is focused on addressing precisely this challenge. A 2026 study protocol proposed evaluating predictive models based on digital phenotyping to determine whether changes in behavior, mobility, and daily routines can be used to build clinically reliable tools capable of anticipating relapse [6].
The direction of research appears increasingly clear. For decades, the information available to psychiatrists and psychologists depended largely on periodic interviews, questionnaires, and in-person consultations. New approaches are incorporating continuous streams of data generated outside the clinic, in the places where people actually live, work, and interact.
There is still insufficient evidence to conclude that smartphones can accurately predict psychiatric relapse. It is also unclear how these systems will ultimately be integrated into clinical practice or what role healthcare professionals will play when algorithms identify early warning signs.
Yet the devices people carry with them every day are beginning to assume an unexpected role. Beyond communication, recording, and connectivity, they may become tools capable of detecting behavioral changes that have historically remained invisible to healthcare systems.
The ubiquitous therapist does not yet exist. But the studies published in recent years suggest that the infrastructure required to build it is already beginning to take shape.
[1] Smartphone Digital Phenotyping in Mental Health Disorders (2025)
[2] Use of Mobile Sensing Data for Longitudinal Monitoring and Prediction of Depression Severity (2025)
[3] Key Features of Digital Phenotyping for Monitoring Mental Health (2025)
[4] Digital Phenotyping Using Smartphones Could Help Steer Mental Health Treatment (2025)
[5] Digital Phenotyping and Digital Monitoring Technologies for Relapse Detection in Mental Health (2026)
[6] Using Smartphone-Based Digital Phenotyping to Predict Relapse (2026)
