Digital mental health provision has expanded enormously, and the quality range within it is very wide.
The categories
Structured therapeutic programmes based on established treatment models, generally with defined modules.
Tools delivering specific techniques — mood tracking, relaxation, breathing.
Platforms connecting users to human therapists.
And general wellbeing applications with no specific therapeutic model.
Which have very different evidence bases and are marketed similarly.
What has evidence
Structured digital programmes based on cognitive behavioural principles have reasonable trial evidence for mild to moderate depression and anxiety.
Which includes programmes recommended in clinical guidelines in several countries.
The critical finding is that guided versions, with some human contact, consistently outperform unguided ones.
The attrition problem
Real-world use shows very high drop-off, with most users stopping within weeks.
Which means trial results, where participants are supported and monitored, overstate real-world effect substantially.
Completion rates in unguided programmes outside trials are low enough to limit population effect regardless of efficacy.
The evidence gap
The large majority of available mental health applications have no published evidence at all.
Which has been documented in reviews of app stores repeatedly.
Claims made in marketing frequently exceed anything tested.
Privacy
Mental health data is among the most sensitive categories.
Investigations have found applications sharing data with third parties including advertisers, in some cases contrary to their own privacy statements.
Which has produced regulatory action in several jurisdictions.
Checking what data is collected, whether it is shared, and whether it can be deleted is worth doing before use.
Regulation
Applications making medical claims may fall within medical device regulation depending on jurisdiction and on the claim.
Which produces a distinction between regulated products and general wellness applications with very different oversight.
Some health systems maintain assessed libraries of applications meeting defined standards, which is a useful filter.
What to look for
Published evidence for the specific product rather than for the general approach.
Whether it is recommended by a health service or assessed by an evaluation programme.
Whether it includes human support.
The privacy policy and what it actually permits.
And whether it has a route to escalate if someone is in crisis.
What they cannot do
Manage risk, respond to crisis or replace assessment.
Which means anyone in distress should contact a doctor or crisis service rather than relying on an application.
They are best understood as a supplement or an interim measure, and the better ones state this themselves.
Conversational agents
Automated systems delivering therapeutic content through dialogue.
Which has expanded rapidly, and the evidence base is early and mixed.
The specific concerns are risk detection, accuracy of responses in crisis, and whether the interaction encourages reliance in a way that delays human help.
Regulatory attention to this category has increased in several jurisdictions.
Access arguments
Digital provision reaches people who cannot access services because of waiting lists, cost, location or reluctance.
Which is the strongest argument for it and is genuine.
The counterargument is that it can become the offer rather than an addition to it, and the distinction depends on how services deploy it.
Digital exclusion
Substantial groups lack the devices, connectivity, skills or confidence for digital provision.
Which overlaps with the populations experiencing the highest need.
Digital-first service design therefore risks widening rather than narrowing access gaps unless alternatives are maintained.
Measurement within tools
Many programmes include standardised symptom measures, which allow progress to be tracked.
Which is genuinely useful and produces data the provider holds.
Understanding what happens to that data is part of deciding whether to use the tool.
Using them sensibly
Choose one with evidence, expect to need some structure to keep going, and treat it as one component rather than as the whole approach.
Which reflects what the trials actually tested — guided programmes with contact, not solitary app use.
If it is not helping after a reasonable period, that is information rather than failure, and it warrants a conversation with a clinician.
Notifications and engagement design
Many applications use engagement techniques developed for consumer products.
Which includes streaks, reminders and gamification, and these increase use rather than necessarily increasing benefit.
A streak that produces guilt when broken is working against the stated purpose of the product.
Free and paid
Free applications generally monetise through subscription upgrades or through data.
Which is worth establishing before entering sensitive information.
Applications provided through health services or non-profits generally have clearer arrangements.
Employer-provided apps
Frequently offered as part of wellbeing packages, and employees reasonably ask what the employer can see.
Which should be nothing individual, with aggregate usage reported at most.
Establishing this explicitly before use is sensible, since the answer determines whether honest use is safe.
Deleting your data
Rights to deletion exist in many jurisdictions and providers must have a process.
Which is worth exercising when you stop using something rather than leaving the account dormant.