Translational markers of adolescent vulnerability to behavioural addictions in the era of generative artificial intelligence

Introduction

The adolescent brain is not just an immature adult brain. Its developmental configuration makes it both extraordinarily plastic and disproportionately susceptible to forming addictive patterns. Dopaminergic reward circuits reach functional maturity years before the prefrontal regulatory networks that govern impulse control and long-term evaluation. The mismatch is neurobiologically determined, and it produces a critical window in which repeated exposure to reward-associated stimuli generates more durable circuit-level change than the same exposure does in adulthood. Generative AI and companion AI applications have entered this window with reinforcement properties no prior consumer technology has combined: variable-ratio conversational reward, anthropomorphic relational dynamics, and effectively unlimited availability.

Aims

To synthesise the translational neuroscience of adolescent addiction vulnerability, identify validated biomarkers predicting behavioural addiction risk, examine emerging evidence on generative AI as an addictive vector for adolescents, and propose a clinically actionable biomarker framework for risk stratification.

Methods

Systematic narrative review (2008–2025) integrating developmental neuroscience, translational addiction markers, and epidemiological and neuroimaging studies of established behavioural addictions. Major datasets: IMAGEN (n=2,000, prospective from age 14), the Adolescent Brain Cognitive Development study (n=11,000), and the mindLAMP and CrossCheck digital phenotyping platforms.

Results

The dual-systems mismatch — striatal maturation outpacing prefrontal completion by up to a decade — has solid grounding in developmental neuroimaging. Three translational markers carry the strongest prospective predictive validity. Steep delay discounting is meta-analytically associated with behavioural and substance addiction risk across 77 studies. The combined profile of high impulsivity and ventral striatal hypoactivation during reward anticipation predicted substance misuse and internet gaming disorder onset in IMAGEN before any disorder developed. Anxious attachment combined with high anxiety sensitivity predicts parasocial relationship formation, which has direct relevance to AI companion engagement. Internet gaming disorder prevalence reaches 10% in adolescent samples under DSM-5 criteria; problematic smartphone use affects 10–23%; habitual social media checking produces measurable changes in reward-sensitive brain networks at one-year follow-up in the ABCD cohort.

Generative AI companion applications present reinforcement schedules of a different order, with reported daily session times exceeding two hours. That is above social media platforms whose adverse adolescent mental health associations are already well established. Qualitative work documents withdrawal-equivalent distress on application unavailability, which meets one core criterion for behavioural addiction. Neuroimaging in internet gaming disorder shows striatal hyperactivation during cue reactivity and reduced dorsolateral prefrontal connectivity, paralleling the neural signature of substance use disorders. We propose a four-domain biomarker panel: behavioural (delay discounting, UPPS-P impulsivity, compulsive use ratings); task-based fMRI (striatal reward anticipation, prefrontal-striatal connectivity); structural MRI (orbitofrontal and dorsolateral prefrontal grey matter volume); and clinical assessment (anxiety sensitivity, attachment style, UCLA Loneliness Scale).

Conclusions

The adolescent brain has a documented neurobiological predisposition to behavioural addiction, and generative AI is the most potent addictive technology to which this generation has been exposed. Prospective research on the interaction between the two is almost entirely absent. The translational base from gaming disorder, social media, and attachment neuroscience is sufficient to design and justify that research now, before the addictive trajectories it would study are fully established.

References

References

1. Casey BJ. et al., 2008. The adolescent brain. Ann. N.Y. Acad. Sci. 1124, 111–126.
2. Laestadius L. et al., 2022. Too human and not human enough: mental health harms from emotional dependence on the social chatbot Replika. New Media Soc. 26(4), 2072–2090.
3. Maza MT. et al., 2023. Association of habitual checking behaviours on social media with longitudinal functional brain development. JAMA Pediatr. 177(2), 160–167.
4. Sohn SY. et al., 2019. Prevalence of problematic smartphone usage and associated mental health outcomes in children and young people. BMC Psychiatry 19(1), 356.
5. Whelan R. et al., 2014. Neuropsychosocial profiles of current and future adolescent alcohol misusers. Nature 512(7513), 185–189.