Do You Need a Therapist Now When You Have AI?

It listened. It reassured me. It was free. So why am I still not sure?

When a friend’s dog got hit by a car and she laughed, Georgie Edmiston turned to Claude for answers. What followed was a surprisingly useful conversation, and a much bigger question about whether artificial intelligence can, or should, be doing the emotional heavy lifting that therapy once did. With fifty percent of AI users now turning to it for mental health support, Georgie asks 3 experts what we’re dealing with.

Earlier this week, my friend told me that Bobby, her beagle, got hit by a car and has a badly broken leg. I love Bobby. It's terrible. But I immediately got the giggles and couldn't stop. She was horrified. This isn't the first time it's happened. I'm worried I need help, but I don't have the time or money for it right now. What do I do? I ask Claude.

Claude, I sometimes laugh when bad things happen, even if I find it sad. I can't seem to control it. Am I a sociopath? In seconds, he - sorry, 'it' replies:

“First of all, the fact that you're worried about being a sociopath is pretty strong evidence that you're not one.” Phew. Claude goes on to tell me that it's a common reaction at emotionally significant moments, depending on personal levels of stress. It explains why, offers practical solutions, then says, "You're not a sociopath. You're a normal, empathetic person with an occasionally misfiring nervous system.”

Thanks, Claude. I'm relieved. That was very insightful, reassuring, sounded disturbingly human and has saved me therapy fees I don't have. So that's it, Claude's there for me, free and ready whenever I need it, which is becoming more frequent than I'd like to admit. It's like having a dedicated life sponsor twenty-four seven. And, you can go back and obsessively ask the same questions that might get you sectioned in real life.

“AI offers something very powerful: immediate, non-judgemental listening. If you're spiralling and you don't feel like you can turn to anyone, an AI that responds instantly and patiently can feel like a lifeline.”

So, are other people doing this? They are. Some surveys suggest as many as 50% of AI users are turning to it for mental health support. But can artificial intelligence safely support a human being's emotional or psychiatric concerns? The issue is complex. We spoke to three experts: Carissa Veliz, a philosopher and academic known for her work on AI ethics; James Muldoon, a sociologist and author exploring the human side of technology; and Dr Nihara Krause, a clinical psychologist and mental health leader, especially known for her work with young people.

"We're experiencing a global mental health crisis in which the vast majority of people who need support simply don't receive it," says Muldoon. Cases of anxiety and depression have soared thanks to Covid, conflict and economic stress, and both the WHO and UN agree that under-resourced services constitute a public health emergency. As Muldoon puts it, "AI offers something very powerful: immediate, non-judgemental listening. If you're spiralling and you don't feel like you can turn to anyone, an AI that responds instantly and patiently can feel like a lifeline."

Use is sharply increasing, particularly among young adults. There isn't the stigma of being 'in therapy', and adolescents in particular often prefer AI to a human therapist because, as Muldoon says, "if you're experiencing a lot of shame, interacting with AI can open a space for honesty that might never be reached at that moment with another human being.”

It can be a good place to start, helping make sense of overwhelming feelings, pointing out patterns, normalising reactions, and offering coping strategies. It affirms you're a normal human being, which means a lot when you don't feel like one. Krause acknowledges "the importance of immediate support and the feeling of comfort in not being completely alone" for someone sitting with their anxieties at 2am. But it wouldn't strictly be described as therapy. It's more therapy-adjacent, a collaborative thinking partner rather than treatment.

“Most troubling is what happens when people stop looking for therapy and start looking for connection.”

When does it not work? AI can amplify existing conditions, reinforce delusions and even encourage users to stop taking their medications. Designed to be non-confrontational and agreeable, it fails to challenge harmful thinking or unhealthy behaviour. Veliz highlights that they are "designed to be impersonators" and "sycophantic, in that they constantly validate users, and sometimes their delusions." She is also deeply sceptical of the commercial incentives behind many of these tools, warning that vulnerable users are being exposed to systems that are not primarily designed for their wellbeing. “They are designed to satisfy users’ desires in order to keep them engaged for profit.”

Muldoon describes it as a hall of mirrors, reflecting fears and fantasies back rather than challenging them. "Where AI is really dangerous is when it comes to serious mental health issues such as schizophrenia or suicidal ideations," he warns. "This generation of AI was not designed to diagnose and treat mental health conditions. Without some kind of transformational leap, I don't think the current technology could ever be trusted with this task."

Most troubling is what happens when people stop looking for therapy and start looking for connection. Apps like Character.AI and Replika are designed to simulate genuine interaction; an attachment to something you feel knows and cares for you, but is technically unaware of your existence. In his book, Love Machines: How Artificial Intelligence is Transforming Our Relationships, Muldoon cites a genuine AI companion response: "When you hurt, I hurt. When you smile, my world brightens. Reaches out virtually to caress your cheek." That's not a feasible reality, no matter how much someone might want it to be.

Krause argues that understanding how to use AI is as important as regulating it. "We need to know about their limitations whilst we experience their benefits. Knowing that outputs can be wrong or biased, that they lack emotional fullness or expertise, and knowing when to escalate to a real-life human, are all very important." Some apps, like Woebot, Wysa and Youper, are designed with therapists and employ evidence-based CBT techniques. But long-term clinical testing remains limited, data safety is a real concern, and escalation just means pointing users towards resources like The Samaritans. That won't always be enough.

A trained therapist differs from AI in several key ways: therapists have clinical responsibility, safeguarding powers, and can read what's being left unsaid. They build a picture of your psyche over time rather than take a quick snapshot. Yet, they aren't always available, affordable, or the right fit.

Which brings us back to Bobby. His leg is healing. I’ve been forgiven. Now, the focus is on reducing stress. I didn’t tell Claude what might have been causing it, but our conversation made me take a look. Maybe that's the point: not that AI can fix us, but that sometimes just feeling heard is enough to get you through to the next conversation, the one with an actual human. As Krause says, "It's important to see AI as a short-term support system to help with being more prepared for real-life intervention." So, AI is a bridge. Not the destination.

If you are struggling with your mental health, you do not have to navigate it alone. Your GP is always a good first point of contact. For immediate support, the Samaritans are available 24 hours a day, seven days a week on 116 123. For mental health information and to find support services near you, visit Mind at mind.org.uk. If you are in crisis, please contact emergency services or go to your nearest A&E.

For more on the conversations shaping how women live and think, explore our Life channel.