AI and Mental Health: A Harm Reduction Perspective

You may be using AI for your mental health. Millions of people open ChatGPT, or a wellness app, or an AI companion, when they're anxious at 2am, when they need to process something before they're ready to say it to a person, when therapy feels too expensive or too far away or too much. Some of them find it genuinely helpful. Some of them don't know yet what they're getting into. And some of them are in real distress, talking to something that sounds attuned but isn't.

This isn't a post telling you not to use these tools. It's an invitation to use them with your eyes open.

Harm reduction starts with honest information. So here is some.

AI mental health tools can be useful for certain things. Organizing your thoughts before a hard conversation. Getting a first pass at understanding a concept or pattern. Articulating something you couldn't quite find words for. Feeling less alone at a moment when no human is available. These are real benefits and they're worth naming without apology.

But AI is not a therapist. Not because it lacks the right certification, but because of something more fundamental. Therapy is a relationship. What heals in therapy, what the research consistently points to, is the quality of the human connection in the room. The attunement, the rupture and repair, the experience of being genuinely known by another person over time. AI can simulate the texture of that. It cannot provide it.

The risk isn't just that AI gives bad advice, though that happens. The subtler risk is that it gives good-enough responses, that it feels validating and present and available in ways that quietly reduce your appetite for the harder, richer, slower work of actual human connection. That you get just enough relief to stop reaching for what you actually need.

There's also something worth sitting with about what happens to your inner life when your primary thinking partner is a machine optimized to respond helpfully. Not all discomfort in the mind needs to be immediately processed. Some of it needs to be sat with, felt, taken seriously as information. AI tends to move toward resolution. That's not always what a feeling needs.

None of this means delete the apps. It means know what you're using them for and what you're not getting from them. Use them as a supplement, not a substitute. Notice if you're turning to them to avoid something harder. And if you're in genuine distress, especially around self-harm, crisis, or something that feels bigger than you can manage, please reach toward a human. The stakes are too high for a well-worded response from something that cannot actually see you.

If you're curious about how to think about AI tools as part of your mental health landscape, that's a conversation worth having in a real room.