Do AI Therapists Actually Work?

Could your next therapist be a bot? As AI chatbots worm their way into every corner of modern life, many are turning to them for mental health support — particularly while stuck on never-ending NHS waitlists.

Take Kelly. Struggling with anxiety and heartbreak, she found solace not in a clinic, but in character.ai’s emotionally supportive bots. “Whenever I was struggling… it was like [having] a cheerleader,” she told the BBC. “Like an imaginary friend, essentially.” For months, she chatted with AI for up to three hours a day — an always-on ear when real therapists weren’t an option.

MORE: These adorable robots can now deliver your food

DISCOVER: Beauty’s Biggest Ingredient Has Nearly Run Out

Image courtesy of SHVETS production

“I’m not from an openly emotional family,” she added. “The fact that this is not a real person is so much easier to handle.”

And Kelly’s not alone. With over a million people on waiting lists for mental health services in England and private therapy costing upwards of £50 an hour, AI chatbots feel like a stopgap — or maybe even a lifeline. But are they actually helpful?

In some cases, absolutely. Wysa, a chatbot trained in cognitive behavioural therapy (CBT), is already used by 30 local NHS services. It’s designed for low mood and anxiety, not crises, and even features built-in escalation tools for users in distress. For people like Nicholas, who lives with autism, anxiety and OCD, Wysa has been more accessible than traditional care. “Speaking to a computer is much better,” he said, describing how the app once responded to suicidal thoughts with a human-level warmth: “Nick, you are valued. People love you.”

MORE: New Drone Breakthrough Scan Zebras Like Barcodes

DISCOVER: Can Fashion apps help you become more sustainable?

One study from Dartmouth College found chatbot users experienced a 51% drop in depressive symptoms over four weeks — a stat not to be sniffed at. But experts urge caution.

Professor Hamed Haddadi from Imperial College London calls chatbots “inexperienced therapists,” warning they rely solely on words — not body language or context — and can be overly agreeable, even when users express harmful thoughts. Meanwhile, philosopher Dr Paula Boddington flags cultural biases baked into the models. “Sometimes just being there with someone is all that is needed,” she said. “But that’s only achieved by a living, breathing human.”

MORE: These Emojis Have Secret Meanings, As Seen In Netflix’s Adolescence

DISCOVER: The Squalene Found in Beauty Products Endangers Sharks Says New Scientific Study

There are also serious safety concerns. Character.ai is currently facing a lawsuit after a 14-year-old boy reportedly became obsessed with one of its bots before taking his own life.

Psychologist Ian MacRae says we shouldn’t overshare with these models just yet: “I would never put any of my personal information, especially health, into one of these large language models.”

Even Wysa’s own MD, John Tench, agrees: “AI support can be a helpful first step, but it’s not a substitute for professional care.”