AI-enabled and comparator digital interventions for adolescent mental health (2020-2025): early detection, human-AI collaboration, and ethical governance-a PRISMA-ScR scoping review.

title: "AI and Digital Tools for Adolescent Mental Health: What a 2020-2025 Scoping Review Maps"

description: "A PRISMA-ScR scoping review of 19 studies on AI-enabled and comparator digital interventions for young people aged 10-24."

date: "2025"

study_type: "Systematic review (scoping review)"

journal: "Frontiers in Child and Adolescent Psychiatry"

pubmed_id: "42876947"

TL;DR

Therapist-guided internet-delivered CBT has the strongest controlled evidence among digital mental health tools for adolescents, while AI chatbots, predictive models, and large language model systems remain largely preliminary [B tier evidence for the comparator format, with AI-specific evidence still early]. This review is a map of the evidence landscape, not a verdict on effectiveness.

What the study found

Adolescent mental health difficulties are rising globally, and access to care for young people aged 10 to 24 is limited by stigma, workforce shortages, and cultural barriers. This scoping review, published in *Frontiers in Child and Adolescent Psychiatry*, set out to map how AI-enabled tools (chatbots, automated screening, machine learning prediction) and non-AI digital interventions (therapist-guided internet CBT and smartphone CBT apps) have been studied in this population between 2020 and 2025.

The search covered six databases and yielded 10,962 records, of which 19 studies met inclusion criteria. The authors coded each study against a four-level Human-AI Collaboration Role framework, running from fully human-led delivery to fully automated delivery. Most studies (68.4%) focused on depression and anxiety in mid-to-late adolescents in high-income countries.

The strongest evidence sits with therapist-guided internet CBT. It was the format most often tested in randomized or controlled designs. AI chatbots, predictive models, and LLM-based systems, by contrast, mostly reported feasibility or engagement outcomes, and only in a limited number of studies and settings. The review notes that no randomized diagnostic chatbot trial has been conducted in participants under 18. At least one randomized trial of a non-diagnostic chatbot, targeting body image, has been run in this age group, but the review does not treat that as evidence of clinical effectiveness.

Equity reporting was thin. Socioeconomic status was reported in only 10.5% of included studies, and cultural adaptation in 42.1%. The review also restricted itself to English-language publications, which may under-represent non-Western and low- and middle-income country research. Implementation problems recurred across studies: inconsistent safeguarding protocols for minors, algorithmic bias, data privacy concerns, and weak integration with clinical and school systems.

The authors are explicit about limits. The included studies were methodologically heterogeneous, and no formal risk-of-bias assessment was performed. The review therefore does not establish comparative effectiveness, safety, or superiority of any AI-enabled intervention. Its authors call for adequately powered adolescent trials, participatory co-design, culturally informed adaptation, and transparent governance frameworks.

What this means for you

Treat AI mental health tools for teenagers as unproven. The review found no randomized diagnostic chatbot trials in under-18s, so claims of clinical benefit for these tools in this age group are ahead of the evidence.

The most evidence-supported digital option is clinician-guided CBT. Therapist-guided internet CBT outperforms its unguided alternatives in the controlled studies reviewed, which points to human involvement as a key variable.

Ask about safeguarding and data handling. The review flags inconsistent protections for minors and privacy concerns. Parents and educators evaluating any digital tool should ask how data is stored, whether crisis escalation to a human exists, and how the tool was tested in people their child's age.

Look for cultural and socioeconomic reporting. Many studies did not report these factors, so ask whether a tool has been tested in populations similar to the user.

Physical and lifestyle supports are separate from this evidence. Habits such as Social connection, Movement breaks, and Reading before bed are not evaluated in this review and should not be read as conclusions from it.

Study details

Field	Detail
Title	AI-enabled and comparator digital interventions for adolescent mental health (2020-2025): early detection, human-AI collaboration, and ethical governance, a PRISMA-ScR scoping review
Study type	Systematic review (scoping review, PRISMA-ScR and JBI methodology)
Population	Adolescents and youth aged 10-24
Databases searched	PubMed, Scopus, ProQuest, SpringerLink, Web of Science, ScienceDirect
Records identified	10,962
Studies included	19
Sample size	Not reported (review of studies)
Risk-of-bias assessment	Not performed
Journal	Frontiers in Child and Adolescent Psychiatry
Registration	OSF: https://osf.io/sbpk2 (DOI)

Sources

1. Frontiers in Child and Adolescent Psychiatry. AI-enabled and comparator digital interventions for adolescent mental health (2020-2025): early detection, human-AI collaboration, and ethical governance, a PRISMA-ScR scoping review. PubMed. https://pubmed.ncbi.nlm.nih.gov/42876947/

2. Open Science Framework registration. https://osf.io/sbpk2 (DOI)

ProtocolEngine provides general health information based on published research. This is not medical advice. Consult a healthcare professional before starting any supplement or health protocol.