Your ChatGPT Therapist

The wins, the losses, and how to navigate the future.

As many have, I recently asked ChatGPT a question I might once have taken to a therapist.

Not for fun. I was curious how it would help me think through a relationship pattern I’ve struggled with before. The response? Surprisingly insightful.

Apparently I’m not alone. A recent survey suggested that around 40% of people are already using AI tools like ChatGPT for therapy or therapy-like support. Whether or not this is true, the number is going to increase.

Let’s unpack what that means. The wins, the losses and the red flags.

Subscribe

The Promise of AI Therapy

At first glance, it’s impressive. There are many ‘wins’ to celebrate. In your pocket, you have a ‘therapist’ that’s:

• Available 24/7
• Free or low-cost
• Trained on millions of therapeutic conversations
• Able to take on any theoretical style you ask for (CBT, psychodynamic, somatic, attachment-focused...)
• Private – it won’t gossip with other humans

And depending on the model and settings, it might even remember what you told it last week.

It can synthesise personal details, recognise patterns, and offer practical, emotionally attuned suggestions — all without judgement, distraction, or awkward silences.

If you’ve never tried it, here’s a prompt to test:

Act as a renowned attachment-based therapist. Help me understand why I feel upset when my partner [insert behaviour]. Ask questions about my past and relationship patterns, and help me work through it constructively — emotionally and practically.

Or just:

Help me deal with feeling upset with [insert person or issue].

The first will give you a better response (prompt design matters). But both can yield surprisingly helpful insights.

In my case, it helped me distinguish several feelings I was having about a pending trip my partner was having. Not only did it help me explore the potential origins of these feelings – for example, in relation to similar situations with my parents long ago – but gave alternative explanations for what was happening and helpful prompts I could use in exploring my feelings with her. It was undoubtedly a useful exercise.

What You Lose When You Gain Convenience

Still, something is lost. Some of it is obvious — like having a real human in the room. Non-verbal communication, nervous system regulation, and eye contact are all powerful components of healing.

A good therapist doesn’t just give advice. They feel you (emotionally!), track subtle shifts in your body and mood, and sometimes challenge you in ways a machine cannot.

The words the AI therapist used were helpful and I internalised the advice. Yet, the ‘therapist’ was unable to use its intuition to pick up on subtle dynamics that occurred. It relied on its words to help me rather than any biological or relational processes such as eye contact, co-regulation and a felt sense of empathy.

Deep therapy often works through relationship dynamics, exploring parallel processes between you and the therapist to learn more about what may not be spoken. You react to your therapist, and they to you.

In psychoanalytic theory, this is called transference and countertransference — unconscious patterns developed from early life relationships that are transferred onto the therapist, and the reaction the therapist has to those transferred emotions, respectively. AI might pick up on some language that signals projection or idealisation, for instance, but it can’t truly engage in that process. It doesn’t have a nervous system, a past, or feelings. And that changes the depth of the process.

But Here’s What’s More Concerning (Red Flags)

The lack of relational depth in AI therapists may be outweighed by the deep intelligence, ease of use and so-called privacy of AI therapists for many. Yet, for me, the deeper concerns aren’t emotional. They’re ethical.

First, AI tools are not neutral. While therapy is also a paid relationship, most human therapists are guided by ethics, standards, and a desire to eventually help you outgrow the need for therapy. AI platforms? Not so much.

Their incentive is often to keep you engaged — because attention and data are monetisable. That doesn’t mean every interaction is harmful. But it introduces a conflict of interest that people should be aware of.

This may foster a sense of dependency on the AI therapist without the usual process of individuation that should be fostered in therapy sessions – potentially damaging for someone with a dependent personality that would benefit from trusting their own judgement instead of others.

Second, what happens to your information?

Depending on the provider and settings, your most private disclosures may be stored, analysed, or in the future used for personalise advertising (many have not ruled this out).

We already see targeted ads from casual searches. Imagine what could be done with the emotional and psychological data we willingly hand over.

The more AI knows about our internal world, the more powerful — and potentially manipulative — it becomes.

So Where Does That Leave Us?

Like all technology, AI therapy offers both opportunity and risk. Whether it becomes a net positive will depend on how we use it — and how critically we think about its role.

Here’s how I see the options emerging:

1. In-person therapy only

Some will continue to see real humans exclusively, out of preference or principle. And honestly, that will always be a valuable path. We are relational beings and if we want to continue to relate to humans, we should practice relating with them, not robots.

2. AI as a helpful assistant or specialised therapist

Others will use ChatGPT and similar tools for day-to-day emotional support, working through relational tensions, or gaining clarity when therapy isn't available. For many, it may function like a reflective journal with a brain. Some will go further and use it for all their therapist needs due to the power and ease of use an AI language model provides. For these people, they should be aware of potential conflicts of interest.

3. Hybrid models

In my opinion, this is the optimal future. AI is here to stay and is undoubtedly a powerful tool. It should augment but not replace person to person therapy. For example, AI tools can support therapists behind the scenes — helping with notes, treatment suggestions, or client reflections. Clients might use AI between sessions or even bring ChatGPT-generated insights to discuss with their therapist. It becomes a tool, not a replacement.

A Rapidly Evolving Frontier

This space is moving fast. Much faster than traditional therapy can adapt.

It will take input from thoughtful clinicians, curious clients, and responsible developers to shape it well. And we’ll need to check in regularly — because what’s safe and ethical today may look different tomorrow.

In the meantime, if you’re experimenting with LLM-based therapy, I’d suggest several things:

1. Drive the ship. The better the guidance, the better the result. As best you can, be specific with the issue, the context and what you are wanting to achieve. Tell the AI therapist the style of therapy you want or if you don’t know, ask it for an overview of evidence based therapies for your specific issue.
2. Know what model you’re using and understand its privacy policy. Part of being a ‘consumer’ in health is your right to informed consent. With AI models, it’s up to us to do the research and be informed. Privacy policies and settings are typically easily found within the platforms or able to be ‘googled’, if you are old school.
3. Notice when it helps — and when it leaves you feeling more confused, dependent, or disembodied. If the therapy is not working or you are feeling like you cannot exist without AI therapy, consider taking a break or finding a human therapist, as some issues will respond best to the emotional messiness of therapy with a human.

What about you?

Have you ever used ChatGPT (or another AI) to help process something emotional?
What worked? What didn’t?

‍