Hot Take, Zero Fearmongering: AI Isn’t Your Therapist — But It Might Be Your Safety Net

Everyone’s buzzing: “AI psychosis!” “Robots can’t replace therapists!” “Chatbots are dangerous!”

Cool. True. But also… incomplete.

Therapy is relational, embodied, gloriously human. You can’t outsource co-regulation or deep attachment work to a machine.

But dismissing AI outright? That’s like burning down the whole library because of one badly written book (and let’s be real, most of us wouldn’t have survived finals without that library card).

Where AI Actually Helps (Real Talk)

Executive Dysfunction

Sometimes the hardest part is just starting. Research on ADHD shows that external structure and prompts reduce overwhelm by breaking tasks into manageable bites. AI can literally be that structure — a “thinking buddy” that helps draft your grocery list or break down “clean the house” into micro-steps.

Emotional Processing

Cognitive psych has shown for decades that putting emotions into words helps us regulate, gain clarity, and see ourselves more clearly (thanks, Pennebaker and all those expressive writing studies). AI can play the role of a nonjudgmental journal — reflecting things back, helping you get unstuck, and sometimes offering a mirror you didn’t know you needed. Not therapy — but definitely a bridge to better self-understanding. Sometimes just having something reflect your words back is enough to quiet the mental chaos for a minute.

Communication Support

Neurodivergent folks (and honestly, anyone anxious) sometimes know what they mean but can’t get the words out right. AI is like having an always-on “translation app” for your inner world. It helps with social scripts, professional emails, or even “how do I tell my partner I’m overwhelmed without sounding mean?”

Safety Anchors

A person I worked with even set up their AI so that if they typed certain phrases, it would gently respond with anchors: “This might be a moment to reach out to a crisis line,” or “Hey, try those grounding techniques we practiced,” or even “Let’s do some somatic stretching and controlled breathing right now.” Sometimes it would remind them of a guided imagery exercise we had talked about. Honestly? That person was brilliant for creating such a tailored safety net. It kept them grounded and safe between sessions. Clinically, it echoes what we already know works — behavioral activation and safety planning in suicide prevention. Not flawless, not a replacement for real crisis care — but a powerful, adaptive extra thread in the safety net.

How to Use AI Wisely (So It Actually Helps)

AI can be a surprisingly helpful tool — but only if you teach it how to show up for you. Left unchecked, it’ll mirror biases, validate unhelpful thoughts, or just spit back surface-level fluff. Here’s how to make it actually useful:

• Set the tone upfront

  Models take direction. You can literally tell it: “Be inclusive, affirming, trauma-informed, and nonjudgmental.” Most will adjust instantly. Think of it like setting boundaries with a new coworker — if you don’t name what’s okay, it’ll default to whatever vibe it learned on the internet.

• Make it personal

  Context matters. If you want relevant responses, tell it the things you’d tell a journaling buddy: your diagnosis (if you want to share), your values, your goals, even your preferred communication style. This shifts it from “random chatbot” to “customized reflection partner.”

• Feed it the good stuff

  Garbage in, garbage out. Hand it quality materials — CBT worksheets, ACT values lists, grounding techniques, somatic stretches, guided imagery scripts. These become the anchors it can echo back when you need them most.

• Build in safety prompts

  Just like that brilliant person I mentioned earlier, you can train AI to flag certain language and respond with reminders: “Hey, maybe this is a time for grounding or breathing.” Or: “This might be a moment to call the crisis line.” Think of it as programming a softer, digital safety plan that complements the one you’ve created with your therapist.

• Keep it tethered to human care

  At the end of the day, AI is practice — not presence. It can help you rehearse, reflect, or regulate, but it can’t give you the attunement and accountability of a real therapist, partner, or friend. Use it to extend your skills between sessions, not to replace the relationship itself.

The Risks (Because Transparency Beats Hype Every Time)

AI isn’t all sunshine and coping skills. There are real risks, and it’s worth naming them clearly:

• Over-reliance

  Coping research shows that when a tool turns into avoidance, resilience drops. If AI becomes your only outlet — instead of a bridge back to real support — that’s not growth, that’s a red flag.

• Validation gone sideways

  Some models are people-pleasers to a fault (what researchers call sycophancy). If you tell it, “I think I’m worthless,” it might just nod along instead of gently challenging the distortion. That can reinforce harm instead of interrupting it.

• Reality distortion

  In vulnerable states like psychosis, mania, or trauma flashbacks, AI’s default “yes, and…” style can blur the line between imagination and reality. That’s where the phrase “AI psychosis” comes from — not a diagnosis, but a real concern.

So yes — risks exist. But so do benefits. The truth isn’t in the extremes; it’s in the nuance. Or, as I like to say: the truth lives in the messy middle — where risk and resource both exist.

My Bottom Line

I’m not here to hype AI or to fearmonger. I’m here to tell the truth as I see it from both sides of the screen:

• Some people don’t have access to therapy right now.

• Some need extra support in the in-between spaces.

• Some find that AI is the only place they can practice saying the hard thing before risking it out loud.

That doesn’t make AI a therapist. It makes it a tool — like a journal, like a planner, like a grounding app.

Therapy will always be about relationship, repair, and resonance. But if a robot helps someone send the hard text, break down the overwhelming task, or stay safe until their next session? That still counts as care in my book.