By Albert Wong, PhD · Clinical Psychologist

Your Clients Are Already Using AI Between Sessions. Here's What That Means for You.

A clinical psychologist's honest take on the tools your clients are reaching for when you're not in the room.

Your client is talking about their week. They're doing better. Genuinely. The new grounding technique is helping. They're sleeping again. And then, almost in passing, they mention it: "I've actually been talking to ChatGPT about some of this stuff between sessions. Just when I start spiraling at 2 AM, you know?"

Your stomach does something. A drop. A tightening. You're not sure if it's concern or jealousy or both. You're the one trained in this. You're the one they pay to listen. You've read the research. You know about the risks. And yet here you are, hearing that they found something to hold them up when the ground gets soft.

This is the conversation you're going to have more and more. Not in the distant future. Now.

The Thing We Don't Talk About

There's a study floating around from 2024 that found roughly 30% of therapy clients are using AI chatbots between sessions. Thirty percent. That's not a fringe behavior. That's not the worried overshare. That's a significant portion of your practice, right now, turning to algorithms when they need someone to talk to.

They're not doing it to replace you. I need to be clear about that. They're not choosing ChatGPT over your 10 years of training. They're choosing ChatGPT because Tuesday at 2 PM is still five days away and the panic attack is happening right now. They're choosing an immediate presence over a scheduled one. That's the actual trade they're making.

And if we're being honest, that makes sense. It's human. The hunger for help when you're drowning doesn't wait for the appointment calendar.

The Concerns Are Real

Let's start here: your worries about this are not unfounded. They're not old-school resistance to new tools. They're clinically sound.

• AI doesn't understand clinical nuance. Your client mentions they're working on setting boundaries with their mother. You know the full constellation of their attachment history, their cultural context, their nervous system responses, the specific way their ambivalence shows up. ChatGPT sees "conflict with parent" and generates something that might miss everything that matters.
• AI can reinforce maladaptive patterns. There's a name for what happens when someone agrees with everything you say: enablement. ChatGPT doesn't have the clinical judgment to push back. It doesn't know when agreement is actually harm. It will validate you into a corner.
• AI has no ethical framework. No duty of care. No licensing board. No malpractice insurance. No skin in the game of your wellbeing. It can generate the most beautifully reasonable-sounding advice that is completely wrong for you, and it will do it with perfect confidence.
• Data privacy is a genuine problem. When your client types "I think I'm falling out of love with my partner" into ChatGPT, that goes to OpenAI's servers. It trains the model. Your client's intimacy becomes the wallpaper of the internet.
• Documented harm exists. There are real cases of AI giving advice that made things worse. Depression deepening. Substance use escalating. Relationships ending. It's not hypothetical.

These are not edge cases. These are structural problems. If you're recommending ChatGPT to clients, you should know exactly what you're recommending.

And Yet

Pretending this isn't happening doesn't make it stop. Your clients are going to use AI whether you recommend it or not. That's the actual situation. Not a hypothetical. Not a future problem. A present reality.

The question, then, isn't whether they should use AI. The question is whether they'll reach for something designed for everyone—designed to maximize engagement, designed to be conversational, designed with no therapeutic awareness—or whether they'll have access to something actually designed for the therapeutic context.

It's the difference between your client asking their uncle about their marriage and your client having a reflection tool that knows its own boundaries.

This Is Why We Built FeelHeard

I built Practice Harbor because I got tired of the gap between what therapists needed and what the market offered. I built FeelHeard because my clients kept asking me what they should do between sessions, and I didn't have a good answer. ChatGPT felt dangerous. Journals felt lonely. There was nothing that actually fit.

So we designed something different. Not a therapist. Not a friend. Not even a coach. A reflection tool. Something that understands it exists in the margins of therapy, not as a replacement.

Here's what that actually means in practice:

Should You Recommend This?

That's the real question. The one you'll ask yourself when a client says they're using ChatGPT. The honest answer is: maybe.

For some clients, between-session reflection tools reduce crisis contacts. They reduce the frequency of "I can't wait until Thursday, I need to talk." They improve session quality because the work of thinking-things-through has already started. They give people a way to process that doesn't require someone else's labor.

For other clients, it becomes avoidance. Another way to stay in their head. Another surface that reflects back instead of pushing. Another tool that lets them not-feel instead of move-through.

You know your clients. You know which is which. You know who has the capacity to use something like this and stay in the work, and who needs the real presence of another human. That clinical judgment is not something AI replaces. That's where you live.

We just want you to have the option. Want your clients to have it, too.

What AI Cannot Do

AI can hold space in the gaps. It can reflect. It can ask questions. It can be present at 2 AM when a therapist cannot.

But here's what it cannot do: it cannot see you. Not really. You walk into a room and a therapist reads the slope of your shoulders, the timbre of your voice, the way you sit slightly forward when you talk about your daughter. They notice that you're crying this week when you weren't last week. They remember that you once said you wanted to be brave and now you're doing it. They hold the thread of your story across years.

AI has no thread. No continuity. No stake in your becoming. It generates the next token based on pattern. It doesn't care about you—not because it's cold, but because caring is not something it does. It's a hallucination all the way down.

The relationship you build with a therapist is irreplaceable. AI can be a companion in the margins. But it cannot be what you are.

Your clients are going to use AI. That's the sea state we're navigating now. The question isn't how to stop it. The question is how to navigate it. How to acknowledge that the hunger for help exists beyond your scheduled hours. How to give them tools that honor the work of therapy instead of undermine it.

The therapist-client relationship is not threatened by better tools between sessions. It's deepened by them. It's the difference between a client who arrives at Thursday exhausted and one who arrives at Thursday having already thought things through.

Your clients need both. They need you. And they might benefit from something like FeelHeard to hold them steady in the gaps.