The way a 12-year-old strings words together — the tiny connectors they use, the rhythm of their sentences — might one day help doctors spot mental health struggles years before they become diagnoses. Researchers at Stanford University have shown that artificial intelligence can analyze how children talk about stressful events and predict with striking accuracy which kids will develop depression, anxiety, or other disorders up to six years later.

The study, published in the journal Nature Mental Health, used four AI language models to examine recorded interviews with more than 200 children between the ages of 9 and 13. The children described stressful experiences in their lives — events ranging from financial hardship and parental divorce to abuse and natural disasters — during sessions that lasted about an hour and a half each.

What the AI found surprised the research team. The style of how children spoke mattered far more than what they actually said. Children who later developed mental health problems tended to construct their sentences differently — using small connector words like "and," "to," and "but" in distinct patterns. These subtle linguistic habits turned out to be stronger predictors than the actual content of the children's stories or even expert assessments of how severe their stressors were.

"Speech is inexpensive and scalable," said Ian Gotlib, a professor of psychology at Stanford who led the study. "It's easy, it's accessible, and it may be a stronger predictor of the development of problems than any of these other factors alone."

Those other factors — things like measuring the stress hormone cortisol, testing how the body reacts to stress, or checking the length of telomeres, which are protective caps on chromosomes that wear down under chronic stress — require blood draws, specialized equipment, or laboratory analysis. By contrast, simply recording and analyzing someone's speech could be done by anyone with a smartphone and an internet connection.

The researchers hope this work opens a door to early intervention. Adolescence is the age when depression and anxiety most often first appear, and once these conditions take hold, they can be notoriously hard to treat. Identifying at-risk children early means families and counselors could offer support before problems become entrenched.

"We believe this study provides a robust proof of concept for the development of scalable tools that identify markers of risk before individuals are diagnosed," said Chase Antonacci, the study's lead author and a doctoral student in neuroscience at Stanford.

The team originally conducted these interviews as part of a long-term project tracking how childhood experiences shape brain development. They had reduced each child's complex interview to a single number representing overall stress exposure, but later wondered what else they might be missing from the recordings — and turned to AI to find out.