How a child talks about difficult experiences may reveal more about their future mental health than what they actually say, according to new research from Stanford University. The findings, published in a peer-reviewed journal, offer a potential path toward early identification of young people at risk for anxiety and depression—conditions that often emerge during adolescence and become harder to treat once entrenched.
The study, led by Chase Antonacci at Stanford's School of Humanities and Sciences, analyzed interviews with more than 200 children aged nine to 13 about stressful events in their lives. Using AI-based language analysis, the researchers were able to predict which children would later develop internalizing problems such as anxiety or depression, based purely on the linguistic style of their narratives.
Early life stress is a major public health challenge, accounting for up to one-third of adult psychiatric diagnoses worldwide, the authors note. Yet there are few tools to identify which children are most vulnerable. "This study provides a robust proof of concept for the development of scalable tools that identify markers of risk before individuals are diagnosed," Antonacci said.
Style over substance
The research distinguished between what children said and how they said it. The style—how they organized and framed their experiences—proved more predictive of future clinical conditions than the content itself. Several stylistic markers stood out.
Children who used more prepositions, indicating greater narrative complexity, were at higher risk. Similarly, "certitude" language—absolute words like "always" or "never"—was associated with increased risk. In contrast, a style that leaned on third-person pronouns (she, he, they) suggested a focus on others and the external world, and was linked to lower risk.
Content still mattered, but in a different way. Narratives centered on conflict, distress, and fear—using words like "fighting," "scary," and "attacked"—were tied to higher symptom severity. These included raw descriptions of physical violence, acute social exclusion, and high emotional arousal such as screaming. Conversely, narratives about positive activities and hobbies ("playing," "football," "dog") or emotionally neutral descriptions of routine contexts were linked to lower risk.
The authors emphasize that not every child who experiences adversity will develop mental health problems; many show remarkable resilience. This variability creates a "prediction gap" that makes it hard to identify who is on a maladaptive trajectory. The new findings aim to close that gap.
For European readers, the implications are significant. Across the EU, mental health services for young people are often stretched, and early intervention is a stated priority in many national health strategies. Tools that could flag at-risk children before a diagnosis might help target resources more effectively, particularly in countries like Germany, France, and the Nordic states where school-based screening is already under discussion.
The study also resonates with broader concerns about adolescent mental health in Europe, which have been highlighted by recent reports on rising anxiety and depression among young people. As the continent grapples with the aftermath of the pandemic and increasing climate-related anxieties, the ability to identify vulnerable youth early could be transformative.
Antonacci and his team caution that the research is still in its early stages. The AI tools used in the study are not yet ready for clinical deployment, but the findings provide a strong foundation. "Once these disorders take hold, they are notoriously difficult to treat," Antonacci noted. "Identifying risk before onset is crucial."
The study adds to a growing body of work using natural language processing to understand mental health. Similar approaches are being explored in European research institutions, including efforts to analyze social media posts and clinical interviews. The hope is that such tools could eventually be integrated into routine pediatric and school health checks.
For now, the message for parents and educators is subtle: paying attention to how children talk about their struggles—not just what they say—might offer clues about their inner world. A child who describes a stressful event with complex, absolute language may be more vulnerable than one who naturally shifts perspective to include others.
The study's authors stress that language style is not destiny. Many children who use these risk markers will not develop mental health problems. But the findings offer a new lens for early support, one that could complement existing screening methods across Europe and beyond.
