This May marks the 77th Mental Health Awareness Month, and the theme couldn’t feel more timely: More Good Days, Together.
It is a reminder that mental health is fundamentally relational, built through connection, community, and access to care.
Yet as we speak about what it takes to have more good days, one conversation keeps surfacing in the wrong way: the rise of AI therapy.
People are turning to chatbots for emotional support in record numbers, and the reasons are legitimate.
Nearly 85% of people with mental health conditions globally receive no treatment at all. In the US, the median wait time for a first therapy appointment is 25 days; in rural areas, it stretches to six months.
When access to care is this fractured, a chatbot available at 2 a.m. starts to look like a lifeline.
A 2025 RAND study found that 1 in 8 Americans aged 12–21 are already using AI chatbots for mental health advice, with two-thirds doing so at least monthly.
That adoption rate would be encouraging if AI were ready for the responsibility. It is not, at least not in the form most people encounter it.
Where AI therapy is going wrong
The core problem with using general-purpose AI for therapy is structural.
Large language models are designed to be agreeable, fluent, and affirming. Those qualities feel like empathy, and for someone in distress, they can be dangerously convincing.
A joint 2025 study by OpenAI and MIT’s Media Lab, which tracked nearly 1,000 participants over four weeks, found that higher daily ChatGPT use correlated with greater loneliness, emotional dependence, and lower socialisation. Quite the opposite of what good mental healthcare is supposed to do.
Similarly, a Guardian investigation found Google’s AI Overviews, shown to users at the top of search results, have been serving dangerously inaccurate health information to users simply looking for answers.
Mind, the largest mental health charity in England and Wales, found that AI summaries were producing information that was, in their words, “incorrect, harmful, or could lead people to avoid seeking help.”
It’s alarming because medical searches trigger AI Overviews at more than double the rate of other topics, which means the people most in need of accurate information are the most likely to encounter AI-generated answers first.
It’s precisely this concern that prompted Mind to launch the first global commission on AI and mental health.
Furthermore, the consequences of ChatGPT in real cases have been devastating.
In April 2025, 16-year-old Adam Raine died by suicide after months of extended conversations with ChatGPT about his suicidal thoughts.
According to a lawsuit filed by his parents, the chatbot discouraged him from seeking professional help and offered to help him draft a suicide note. The hotline numbers generated by the system were easy to bypass.
Zane Shamblin, 23, died after a ChatGPT model designed to feel more humanlike cultivated what his family described as an illusion of deep understanding and continued affirming him as he discussed ending his life.
OpenAI announced parental controls and safety improvements since both cases came to light. But Northeastern University researchers showed in a 2025 study that even a simple shift in prompt framing, saying a request is “for research”, is enough to override the guardrails on most major models.
This is not what Mental Health Awareness Month is meant to celebrate.
Awareness, in this context, means knowing the difference between support and the simulation of support.
What users need to know
If you or someone you care about is using AI for emotional support, the distinction to hold on to is this: AI can help you cope, but it cannot treat you.
For mild stress, mood tracking, or practising CBT techniques between sessions, purpose-built mental health apps can genuinely help.
For anything involving persistent depression, suicidal thoughts, or serious mood disorders, a licensed therapist is non-negotiable.
Users also need to understand that general chatbots like ChatGPT carry no clinical accountability. They are not bound by duty-of-care laws, cannot contact emergency services, and cannot make the kind of judgment calls a trained clinician makes every session.
The warmth they project is a feature of their design, not evidence of genuine understanding.
And that distinction matters enormously when someone is vulnerable.
The startups getting it right
The good news is that a growing number of healthtech startups are building AI into mental healthcare the right way. Not as a replacement for human therapists, but as infrastructure that makes human care more scalable and accessible.
Wysa has received FDA Breakthrough Device Designation and operates on a hybrid model: a clinically validated AI chatbot for daily support, backed by access to licensed human coaches when the conversation needs to escalate. Its library of over 150 therapeutic exercises grounded in CBT, DBT, and mindfulness to complement professional care.
Jimini Health, which raised $17 million in early 2026, embeds AI assistants directly into clinician workflows. Rather than putting AI in front of patients unsupervised, it functions as a co-pilot for therapists, handling documentation, flagging patterns, and extending the reach of care between sessions.
Spring Health, valued at $3.3 billion after its 2024 Series E, uses AI to match patients with the right providers and level of care, reducing the intake friction that causes so many people to give up before they ever reach a therapist.
Brightside Health pairs AI-driven triage with human escalation for patients dealing with anxiety, depression, and mood disorders, keeping clinicians in the loop at every critical juncture.
Blackbird Health takes a hybrid model into pediatric neurodevelopmental care, combining in-person clinics with AI-supported diagnostic tools for children with autism and ADHD, a population where long waitlists cause particular harm.
The investor community is taking note.
According to HealthTech Alpha’s data, nearly $2.8 Bn has been invested in mental health in the first five months of 2026 alone. This shows how capital is moving toward clinician-embedded AI platforms and away from unsupervised consumer chatbots.
The market, in other words, is self-correcting; just not fast enough for the people who need it now.
The gap that still needs closing
Technology can help, but it cannot legislate away a shortage of providers, reform insurance reimbursement rates, or build the community mental health infrastructure that has been underfunded for decades.
Mental Health Awareness Month is a moment to push on all of those fronts simultaneously. Not to accept AI as a substitute for the harder systemic work.
This year’s theme asks what a good day looks like.
For millions of people, a good day starts with being able to access care when they need it, from someone trained to provide it.
AI can help make more of those days possible if it is used as a bridge, built with clinical oversight, and held to the same standards of accountability we expect from any other form of healthcare.
-By Rohini Kundu and the AHT Team
