Digital delivery is a format, not a treatment theory
A meditation app, mood diary, peer forum, cognitive behavioural programme, therapist video visit and prescription software are different interventions. Evidence for one should not be generalized to the others. WHO estimates that depression affects 5.7% of adults globally, including 6.9% of women and 4.6% of men, and notes that effective treatments exist. [1] Digital delivery may expand reach, but it does not remove the need for diagnosis, choice, safety and follow-up.
NICE includes guided self-help among options for less severe depression. It describes structured CBT, behavioural activation, problem-solving or psychoeducation delivered through print or digital materials with support from a trained practitioner, usually over six to eight sessions. Access, motivation, computer ability and reduced capacity for individual adaptation should be considered. [2]
- Name the therapeutic model and level of human support.
- Distinguish symptom tracking from treatment.
- Do not present a wellness tool as a substitute for indicated clinical care.
Internet CBT has comparative evidence
A 2021 individual-participant network meta-analysis identified 42 eligible randomized trials and synthesized data from 8,107 participants. Guided and unguided internet CBT were more effective than controls for depression. Guided treatment produced a small additional post-treatment improvement, with a mean PHQ-9 difference of 0.8 points, and the advantage was greater for people with more severe baseline symptoms. [3]
A separate component network meta-analysis included 76 trials and 17,521 participants with reported sex, 71% of whom were women. It found that combining human and automated encouragement reduced dropout, while effects of individual programme components varied and uncertainty remained. [4] These results support structured interventions, not every app carrying a CBT label.
- Look for randomized evidence on the exact programme.
- Check completion, deterioration and adverse events alongside average symptom change.
- Match support intensity to severity, complexity and preference.
Assessment and escalation remain clinical responsibilities
Digital entry should begin with a proportionate assessment of symptoms, functioning, safety, bipolar history, substance use, trauma, medical contributors and current treatment. A score is not a complete diagnosis. Suicidal thoughts, psychosis, mania, inability to care for oneself or immediate danger require urgent human assessment under local emergency pathways.
Every programme should state response times, crisis limitations and what happens when symptoms worsen or a person stops engaging. Automated language detection can support triage but should not be the only safety layer. NICE recommends shared decisions that consider clinical needs and preferences, and more intensive or specialist care for severe, psychotic, chronic or treatment-resistant presentations. [2]
- Screen for bipolar disorder before treating presumed depression when clinically indicated.
- Do not rely on asynchronous messaging for emergencies.
- Confirm that referrals are completed, not merely generated.
- Tell users when a channel is not continuously monitored.
Women's mental health needs contextual care
Symptoms may interact with menstruation, fertility treatment, pregnancy, postpartum recovery, menopause, pain, caregiving, violence or medication changes. These contexts do not make depression or anxiety less real, but they may affect differential diagnosis, treatment choice and urgency. ACOG recommends screening for depression and anxiety at the initial prenatal visit, later in pregnancy and postpartum, with systems for assessment, treatment and follow-up. [5]
WHO reports that almost one in five women experiences a mental health condition during pregnancy or the year after birth. [6] Digital pathways can support repeated screening and access, but postpartum psychosis, suicidality and severe functional decline require immediate specialist or emergency care. Medication decisions during pregnancy or breastfeeding need individualized clinical discussion rather than blanket reassurance or prohibition.
- Record reproductive stage and relevant treatment changes.
- Offer trauma-informed and culturally responsive choices.
- Include sleep, pain, safety and caregiving constraints in the plan.
Equity and privacy shape real effectiveness
A programme cannot be effective for someone who lacks a private device, stable connection, language access or confidence using it. Outcomes should be monitored by relevant demographic and access groups, including who starts, completes, deteriorates and transfers to human care. Alternatives such as telephone, in-person care or supported access should remain available. [2][6]
Mental health data require clear boundaries. Users should know who can read messages, whether data train algorithms, how long records are retained and how information enters the clinical record. Evaluation should include symptom change, functioning, therapeutic alliance, crisis events, dropout, equity, workload and cost. Digital care is strongest when it expands a stepped-care system rather than becoming a low-cost dead end. [2][6][7]
- Can a user reach a qualified person when needed?
- Are outcomes reported beyond engagement and satisfaction?
- Does the service have a safe handoff to local care?
- Can people choose a nondigital alternative?
What the evidence cannot yet answer
- Meta-analysis results apply to studied structured interventions, not all mental health apps.
- Trial participants may have more support and fewer complex conditions than routine-care users.
- Average benefits do not predict an individual's response.
- Evidence for crisis chatbots and generative AI therapy remains much less established than evidence for structured internet CBT.
Questions worth taking into care
- What evidence supports this exact intervention and population?
- How much qualified human support is included?
- What is the crisis and deterioration pathway?
- Are reproductive stage, trauma, safety and medical contributors considered?
- Can users understand and control how sensitive messages are used?
Source record
Evidence used in this review
Sources were selected for clinical authority, methodological relevance and traceability. Links open the original guidance, public-health record or research publication.
- [1]Depressive disorder
World Health Organization · 2025
- [2]Depression in adults: treatment and management
National Institute for Health and Care Excellence · 2022
- [3]Internet-Based Cognitive Behavioral Therapy for Depression
JAMA Psychiatry · 2021
- [4]Dismantling, optimising, and personalising internet cognitive behavioural therapy for depression
The Lancet Psychiatry · 2021
- [5]Screening and Diagnosis of Mental Health Conditions During Pregnancy and Postpartum
American College of Obstetricians and Gynecologists · 2023
- [6]Guide for integration of perinatal mental health in maternal and child health services
World Health Organization · 2022
- [7]NIST Privacy Framework
National Institute of Standards and Technology · 2020
This evidence synthesis is for general information. It does not diagnose a condition or replace care from a qualified health professional. Treatment choices depend on individual history, examination, local guidance and informed preference. Emergency or rapidly worsening symptoms need urgent local medical assessment.