I’ve been watching the mental health app space explode over the past two years, and honestly, it’s equal parts exciting and concerning. On one hand, we’ve got more tools than ever to support mental wellbeing. On the other, there’s a lot of slick marketing disguising apps that have zero clinical evidence behind them.

So I’ve spent the past few weeks digging into which AI-powered mental health apps available to Australians actually have research backing them — and which ones are, well, just chatbots with a calming colour palette.

The Ones With Genuine Evidence

Let’s start with the good news. A handful of AI-driven mental health tools have earned their stripes through proper clinical trials.

Woebot remains the gold standard. Developed out of Stanford, Woebot uses conversational AI grounded in cognitive behavioural therapy (CBT) principles. It’s backed by multiple peer-reviewed studies including randomised controlled trials showing significant reductions in depression and anxiety symptoms. The app adapts its approach based on your responses, and it’s honest about what it can and can’t do. It won’t pretend to be your therapist. It’s a support tool, and it’s clear about that.

Wysa is another strong contender. The app has published clinical evidence showing effectiveness for managing anxiety, depression, and chronic pain. It’s been adopted by several Australian employers as part of employee assistance programs, and it’s got endorsement from the NHS in the UK — which, say what you will about the NHS, they don’t endorse things without rigorous evidence review.

MindSpot, while not strictly an AI app, deserves mention because it’s a free Australian-developed program from Macquarie University that uses guided digital therapy with clinician oversight. It’s funded by the Australian Government and has strong outcome data across thousands of patients. If you’re looking for something with local clinical validation, this is probably your best starting point.

The Grey Area

Then there’s a middle tier — apps using AI that show promise but haven’t yet built a robust evidence base.

Replika and similar AI companion apps are interesting. Some users report genuine emotional benefit from having an always-available conversational partner. But “users feel better” isn’t the same as clinical evidence of therapeutic effectiveness. These apps occupy a weird space between social connection and therapy, and the research hasn’t caught up with the claims.

Several Australian startups are building AI-powered mood tracking tools that claim to predict depressive episodes before they happen. The technology behind this — using patterns in typing speed, app usage, and voice tone — is real and being studied in academic settings. But we’re still in early research phases. Claiming predictive mental health monitoring as a consumer product feels premature.

The Ones That Worry Me

Here’s where I get a bit fired up. The App Store and Google Play are full of apps slapping “AI-powered therapy” onto what is essentially a scripted decision tree with some GPT-generated responses bolted on.

Some red flags to watch for:

• No published research. If an app claims to treat or manage mental health conditions but can’t point you to a single peer-reviewed study, that’s a problem.
• No clinical oversight. Good mental health apps have clinical psychologists or psychiatrists on their advisory boards — not just “wellness coaches.”
• Vague claims. “Our AI understands you” tells you nothing. What therapeutic framework does it use? What training data? What safeguards for crisis situations?
• No crisis protocols. Any mental health app that doesn’t have clear pathways to human support when someone expresses suicidal ideation is genuinely dangerous. Full stop.

The Black Dog Institute has been vocal about the need for regulation in this space, and I think they’re right. Australia’s TGA regulates medical devices, but most mental health apps fall through the cracks by positioning themselves as “wellness” rather than “health” tools.

The Ethics of AI in Mental Health

There’s a broader question here about how AI should be built for mental health applications. It’s not enough to have a working model — the model needs to be safe, private, and transparent.

Firms like Team 400, which focus on building ethical health AI, argue that mental health applications need fundamentally different safeguards than other AI products. I think that’s right. When someone is in a vulnerable state and talking to an AI, the consequences of a bad response aren’t just inconvenience — they can be genuinely harmful.

Data privacy is another concern. Mental health data is among the most sensitive information a person can share. Several popular apps have been caught sharing data with advertisers or using conversations to train models without clear consent. Before you download anything, read the privacy policy. I know, I know — nobody reads those. But for mental health apps, please make the exception.

What I’d Actually Recommend

If you’re looking for AI-supported mental health tools right now, here’s my honest advice:

1. Start with MindSpot or Head to Health. Free, Australian, evidence-based, clinician-supported. No AI hype, just solid digital therapy.

2. If you want an AI chatbot, use Woebot or Wysa. They’ve done the research. They’re transparent about limitations. They have crisis protocols.

3. Don’t replace human therapy with an app. These tools work best as supplements — between sessions, during waiting lists, or for mild symptoms. If you’re experiencing moderate to severe mental health issues, you need a human professional.

4. Be sceptical of anything that sounds too good. An app that claims to “cure anxiety with AI” is lying to you. Mental health isn’t a bug that gets patched with software.

The technology is genuinely promising. AI can make evidence-based therapy techniques more accessible, more affordable, and available at 2am when you can’t sleep and your therapist’s office is closed. But we’re in a Wild West phase right now, and the gap between what’s marketed and what’s proven is enormous.

We deserve better than wellness theatre dressed up as technology. And the good news is that the genuinely effective tools do exist — you just have to know where to look.