SWITCHED ON
The daily technology series nobody asked for but everyone needed
Mind the App: Technology and the Mental Health Question
There are more mental health apps than there are regulated mental health interventions with solid clinical evidence. This fact is not a coincidence and it is not benign. But the underlying question — can technology help? — deserves a more nuanced answer than either the app stores or the sceptics are providing.
The global mental health treatment gap — the proportion of people with a diagnosable mental health condition who receive no treatment — is estimated at between fifty and eighty percent depending on the country and condition. Any technology that meaningfully closes that gap is doing something important. The question is whether the technology being deployed is actually doing it.
Yesterday we went underwater — deep-sea mining's irreversible ecological stakes, offshore wind's genuine success story, ocean carbon capture's early research status, and the governance gap that is the recurring villain of this series applied to the last major unexploited frontier on Earth. Today we are surfacing and going somewhere considerably more intimate. Mental health technology: the apps, chatbots, digital therapeutics, and AI-powered tools that have proliferated in response to a genuine and serious global mental health crisis, and that exist in a space between meaningful innovation and exploitative wellness theatre that is difficult to navigate without a guide. We are going to try to be that guide. This one matters.
01 — The Actual Problem Being Addressed
The global mental health crisis is real and the numbers are staggering. The World Health Organisation estimates that depression and anxiety disorders affect more than 970 million people globally. Suicide remains one of the leading causes of death in people under 35. The economic cost of mental health conditions — in lost productivity, healthcare utilisation, and social welfare — is estimated in the trillions of dollars annually. And the treatment gap — the proportion of people who have a condition and receive no treatment — is extraordinary: in high-income countries, roughly fifty percent of people with a mental health condition go untreated. In low- and middle-income countries, the figure is closer to eighty percent.
The reasons for the treatment gap are multiple and interacting. There are not enough mental health professionals. The ones that exist are concentrated in wealthy urban areas. Therapy is expensive and in most countries inadequately covered by insurance or public health systems. The stigma associated with mental health conditions remains significant in many cultural contexts, deterring people from seeking help even when it is available. Wait times for public mental health services in countries with nationalised healthcare run to months or years for non-crisis care. These are not problems that technology created. They are problems that technology is being asked to help address, in a context where the demand for mental health support has never been higher and the supply of qualified human practitioners has never been sufficient to meet it.
02 — What Digital Therapeutics and Mental Health Apps Actually Are
The mental health technology market is a spectrum from the genuinely rigorous to the thoroughly irresponsible, and the terminology does not reliably distinguish between them. Digital therapeutics — a specific term for software-based interventions that have been validated in clinical trials and in some cases received regulatory approval from bodies like the FDA — sit at the rigorous end. Mental wellness apps — a far larger category that includes everything from guided meditation to mood diaries to AI chatbots — occupy a much wider and more variable middle ground, often with minimal or no clinical evidence of efficacy.
The best-evidenced digital mental health interventions are primarily digitised versions of established therapeutic modalities — computerised cognitive behavioural therapy for depression and anxiety has a reasonable evidence base, with effect sizes in well-conducted trials that are modest but consistent, roughly comparable to face-to-face CBT for mild to moderate conditions. Programmes like Beating the Blues and MindSpot have clinical trial evidence supporting their use as part of stepped care models, where digital intervention serves as a first-line option for people with mild conditions while more intensive human support is reserved for those with more severe presentations.
The category that has attracted the most attention and the most concern is AI chatbots for emotional support and therapeutic conversation. Woebot, Wysa, and a proliferation of general-purpose AI models positioned as mental health support tools represent a genuinely novel category — accessible twenty-four hours a day, infinitely patient, stigma-free, and available at a price point (often free or low-cost) that qualified human therapy is not. The evidence for their efficacy is considerably more limited than their adoption suggests.
An AI that is available at 3am when you are in distress and has no judgment and infinite patience is providing something genuinely valuable. Whether what it is providing constitutes treatment, support, entertainment, or a substitute for care that would serve the person better is a question the evidence has not yet cleanly answered — and that the companies providing the tools have a commercial interest in not pressing too hard.
03 — The Evidence Problem
A 2022 analysis of mental health apps available on major app stores found that the vast majority had no published clinical evidence supporting their claimed benefits. Of those that did have some evidence, the majority was from small, industry-funded trials with limited follow-up periods, measuring intermediate outcomes like symptom self-report rather than clinically meaningful endpoints like diagnosis remission or functional improvement. This is not a situation of honest uncertainty about a new technology's effects. It is a situation in which products are being used by millions of people for serious health conditions, on the basis of marketing that significantly outpaces the evidence supporting it.
The regulatory situation compounds this. Mental health apps exist in a grey zone. Those that make specific therapeutic claims require regulatory approval in the US and EU. The majority of apps are careful to position themselves as wellness tools or emotional support rather than medical devices — a framing that allows them to avoid the clinical evidence requirements that regulated digital therapeutics face, while still being used by people who are, in practice, substituting them for clinical care they are not receiving. The FDA's digital health framework and the EU's medical device regulations both address this category to some extent, but the pace of app development and the difficulty of distinguishing wellness from therapy in practice has produced a regulatory environment in which the most rigorously validated tools face higher barriers than the least validated ones.
04 — The Privacy Dimension
Mental health data is among the most sensitive personal information that exists. Disclosures made in the context of seeking mental health support — details about trauma, relationships, substance use, suicidal thoughts, psychiatric diagnoses — carry significant potential for harm if they reach employers, insurers, governments, or other parties the person did not intend to receive them. The data privacy practices of mental health apps have been the subject of consistent and disturbing scrutiny.
A 2021 investigation by Mozilla rated twenty-eight popular mental health apps on privacy practices and found that the majority shared data with third parties including Facebook and Google, often for advertising purposes, and that the privacy policies were typically written to permit extremely broad data use. BetterHelp, one of the largest online therapy platforms, paid a $7.8 million settlement to the US Federal Trade Commission in 2023 after the FTC found it had shared users' sensitive mental health information with Facebook and Snapchat for advertising targeting purposes — data that users had shared specifically in the context of seeking mental health treatment. The company had informed users that their data would not be shared for marketing purposes. The FTC found this claim was false.
The combination of sensitive data, inadequate privacy practices, and a regulatory environment that has been slow to apply health data protections to the app category creates a situation in which seeking help for a mental health condition through a digital tool may simultaneously provide some benefit and generate a data trail that could be used against the person who sought that help in contexts including employment, insurance, and legal proceedings. This is not a theoretical risk. It is a documented reality that disproportionately affects the people who are most vulnerable — those who turn to low-cost digital tools because they cannot afford or access regulated clinical care.
05 — What Good Technology Can Actually Do
None of this is an argument that technology has no role in mental health. It is an argument for a specific and honest accounting of what role it can and cannot play, and for regulating the space to ensure that what is being provided is what is claimed.
Technology is genuinely well-suited to several specific mental health applications. Reducing access barriers — making support available outside of business hours, without a waiting list, at low or no cost, without requiring physical proximity to a practitioner — is a real contribution to the treatment gap problem. Crisis detection — using passive data from smartphones and wearables to detect early warning signs of depressive episodes or psychotic breaks in people with known conditions, enabling earlier intervention — is a legitimate and evidence-developing application. Augmenting human therapists — using AI to automate administrative burden, flag session content for clinical review, or extend therapeutic contact between sessions through structured exercises — has more clinical logic than replacing therapists entirely. Peer support platforms, which connect people with shared experiences, have a documented evidence base for specific conditions including eating disorders and addiction recovery.
What technology is less well-suited to is replacing the therapeutic relationship entirely for people with moderate to severe conditions — not because the technology is bad but because the therapeutic relationship itself appears to be a significant active ingredient of effective psychotherapy, not merely a delivery mechanism for techniques that could equally be delivered by software. The evidence on this is not conclusive in every direction, but it is sufficient to counsel caution about the strongest claims that AI therapy can fully replicate what trained human clinicians do for people with serious mental health needs.
The right framing for technology's role in mental health is as one component of a system that must also include adequate investment in human mental health services, insurance coverage reform, stigma reduction, and the social determinants of mental health — housing, economic security, social connection — that no app can address. Technology at its best extends the reach of a system that is adequately resourced. At its worst it provides a justification for not adequately resourcing that system, because something is now available for free in the app store.
Tomorrow we are going somewhere that brings together threads from across this entire season — the internet of things, smart homes, and the surveillance architecture being built into the places where people feel most private. Connected devices, data practices, the security vulnerabilities that researchers keep finding and manufacturers keep not fixing, and the question of what it means to live in a home that is permanently online. See you then.
Switched On is a daily technology series covering the ideas, systems, and arguments shaping the digital world. Opinionated. Witty. Occasionally wrong. Always worth the argument.
