Researchers have found that the way children speak about stressful experiences may help predict their future mental health more accurately than expert clinical assessments. In a study published in Nature Mental Health, scientists used four natural language processing (NLP) models to analyse recorded interviews with more than 200 children aged 9 to 13. The models accurately predicted which children would develop mental health disorders up to six years later, suggesting that speech analysis could become a practical, scalable tool for identifying young people at risk before symptoms appear.
The study found that how children spoke was more informative than what they said. Rather than focusing on the specific stressful events children described, the AI models detected patterns in sentence construction and word choice, including the use of common connector words such as and, to, and but. These subtle features of language consistently outperformed the actual content of children’s stories in predicting later mental health challenges. “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 lead author Chase Antonacci, a doctoral student in neuroscience at Stanford University.
Adolescence is the period when conditions such as depression and anxiety most commonly emerge. Yet, clinicians currently have limited ways to identify which children are most likely to develop these disorders. Existing approaches often rely on time-intensive clinical assessments or biological measures, such as cortisol levels, stress responses, or telomere length, which require specialised equipment or laboratory testing. “Speech is inexpensive and scalable,” said senior author Ian Gotlib, professor of psychology at Stanford. “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.”
The research drew on interviews collected as part of a long-term study examining how early-life stress influences brain development and mental health. Each child participated in a 90-minute interview covering stressful experiences, which were originally evaluated by experts using the Traumatic Events Screening Inventory (TESI). The experts assigned each child a single score representing cumulative stress exposure. However, the researchers suspected that reducing complex conversations to a single number overlooked valuable information contained in the children’s language.
To test this idea, the team applied four NLP models that have previously been used to detect mental health signals in adults’ writing. The analyses revealed that linguistic style was the strongest predictor of future mental health. Certain language patterns, including frequent use of first-person pronouns, prepositions, and conjunctions, have previously been linked to mental health conditions. Although speech content was less predictive, children who described severe physical violence or social exclusion were more likely to experience later mental health problems. In contrast, references to supportive relationships, sports, school clubs, and even therapists or counsellors were associated with greater resilience.
The researchers emphasise that larger studies are needed before language-based screening tools can be introduced into clinical practice. Nevertheless, the findings suggest that analysing everyday speech could provide a simple, low-cost way to identify children who may benefit from early support. If confirmed in future research, the approach could eventually allow clinicians to analyse children’s speech recordings—potentially even those collected on smartphones—to detect mental health risks years before disorders develop, creating new opportunities for early intervention and prevention.
More information: Chase Antonacci et al, Natural language processing of youth speech predicts psychopathology across adolescence, Nature Mental Health. DOI: 10.1038/s44220-026-00683-9
Journal information: Nature Mental Health Provided by Stanford University
