What Your Child's Words Could Reveal About Their Future Anxiety Risk
By Stuart Thompson | The STILL Method
A mum I worked with a few years ago used to tell me her son was "fine, he just doesn't talk about it much." He'd been through a difficult house move and the loss of a grandparent within the same year. On paper, nothing stood out. He wasn't withdrawn, he wasn't acting out, and if you'd handed him a standard risk checklist, he'd have scored low. It was only in how he described those events, in the tightness of his language, the certainty in his voice, the way everything came back to himself, that you could sense something was being carried rather than processed.
New research suggests that instinct may have more scientific backing than we realised.
In short
- A new NIH-funded study (Stanford University, published in Nature Mental Health, July 2026) analysed recorded interviews with 204 children aged 9 to 13 about stressful life events.
- AI language analysis found that how a child talked about stress, not just what they described, predicted the onset of anxiety and depression up to six years later.
- Rigid, absolute language, heavy use of "I", and complex, rumination-style sentence structure were strong risk markers. Talk of hobbies, routines, and support networks looked protective.
- The model outperformed standard clinical risk assessments, though it isn't a diagnostic tool and isn't available for use yet.
What the researchers actually found
The study, led by Ian H. Gotlib's team at Stanford and supported by the National Institute of Mental Health, followed 204 children who'd taken part in structured interviews about early life stress. None had a mental health diagnosis at the time of the interview. Researchers then checked back in four and six years later to see who had gone on to develop depression or an anxiety disorder.
Rather than analysing what the children said about their experiences, the team used natural language processing to look at the mechanics of how they said it: sentence structure, word choice, grammar, and recurring themes. Four separate language models were tested against the usual gold standard, expert-rated stress severity scores, and the language models proved more accurate at predicting later anxiety and depression than the traditional clinical risk ratings.
The strongest predictors weren't dramatic. They were subtle patterns: greater use of function words like prepositions (associated with more complex, potentially ruminative narrative style), language that expressed things in absolute, fixed terms, and a heavy reliance on first-person singular pronouns. Children whose stories leaned on routine, structured activity, and access to support tended to fare better. Those whose accounts centred on physical violence or social exclusion were at greater risk.
Why this matters beyond the lab
We talk a lot in The STILL Method about anxiety as something the nervous system learns over time rather than something people simply have. While this study wasn't testing that theory directly, its findings fit comfortably with that way of thinking. It suggests that long before obvious symptoms appear, children may already be organising stressful experiences in ways that reflect how their nervous system is adapting to them.
That's a genuinely useful reframe for anyone working with children, whether as a parent, teacher, or coach. It shifts the question away from "has something bad happened to this child" and towards "how is this child organising what happened to them in their own mind." Two children can go through the same house move or bereavement and come out with very different internal narratives, and it's that narrative, not the event itself, that this research links to future risk.
It's worth being clear about what this study doesn't offer. There's no tool a parent or practitioner can use today to run this kind of analysis, and the researchers themselves describe it as an early step, not a finished diagnostic. What it does offer is a good reason to pay closer attention to the shape of a child's language around stress, not just its content, and not to be reassured just because a child seems calm on the surface.
What to listen for
You don't need an AI model to start noticing some of these patterns. When a child talks about something difficult, it's worth listening for whether their account is rigid ("it was terrible, it's always terrible"), heavily self-focused, or unusually elaborate and looping. It's also worth noticing the opposite: children who can place a hard event alongside ordinary things, their football team, their friends, a routine they still have, tend to be describing a nervous system that's finding its feet again rather than one that's stuck.
None of this is about diagnosing a child from a conversation. It's about listening differently: less for the facts of what happened, more for how settled or unsettled the telling of it sounds.
One thing I'd be careful of is assuming that quieter language is healthier language. Some children don't tell a looping story because they've processed what happened. They stop telling it because they've decided nobody is listening, or because avoiding it feels safer. The goal isn't to analyse every sentence. It's to stay curious about how a child is making sense of their world.
Frequently asked questions
Can this research actually predict which children will develop anxiety or depression?
It showed strong predictive results in a research setting, outperforming standard clinical risk scores, but it isn't an available clinical tool. It's an early, promising finding rather than something parents or schools can use directly yet.
Does this mean my child's exact words are a diagnosis?
No. The study looked at broad language patterns across many children over years, not single conversations. One rigid sentence from a child isn't a red flag on its own.
What should I do if I recognise these patterns in my child?
Treat it as a prompt to stay curious and connected, not to panic. If you're genuinely concerned about a child's wellbeing, a GP, school counsellor, or qualified therapist is the right next step for assessment.
How does this connect to nervous system regulation?
It supports the view that anxiety and depression risk often builds gradually, through how the nervous system processes and narrates stress over time, rather than appearing suddenly or being fixed from birth. That's the same thinking behind our nervous system resource hub.
Where to go from here
If this kind of research interests you, whether as a parent trying to understand a child better or a practitioner wanting to work more effectively with anxiety, our nervous system hub brings together the STILL Method's approach to anxiety as a trainable nervous system state. If you're considering training to support children and families through anxiety professionally, our anxiety coach training is a good place to start.
About the author
Stuart Thompson is the founder of The STILL Method, an accredited coach training organisation, and has worked in anxiety and emotional health for over 25 years, including as a social worker. He holds an ACCPH Chartered Fellowship and is a GHR Senior Member (SQHP), and is the author of 90 Days With Your Nervous System: Not Against It. His work has been featured in The Guardian and on the BBC.
