Why see a psychotherapist when I can talk to Claude for free?
- Amanda Bradley
- 3 hours ago
- 9 min read
There's a reason I work with people. Technology and I do not get along. But, I am willing to recognise that technology is here to stay. We talk to our AI-powered computers and they “answer”. AI agents have names, voices and humour. Talking with them feels "real". The Turing test has been passed.
So why see a psychotherapist when you can just ask Claude?
When I was first asked this question, my initial response was one of rapid dismissal. “Surely you’ll be able to tell it’s giving stock responses”. I discovered this wasn't necessarily the case when I recently asked Claude for some help with my ADHD.
Case Study: Claude and the ADHD shutdown
I was diagnosed with ADHD earlier this year. Last month, I went into ADHD shutdown in the middle of a supermarket. It was my first such experience since diagnosis. I had had similar occurences on and off for many years. Fifteen years ago, a Harley Street neurologist diagnosed me with “a rare form of migraine" and recommended lying down in a dark room. This of course worked because after this type of mental paralysis, recovery begins with sensory deprivation, more commonly known as"lying down in a dark room".
As this was the first time I’d considered these symptoms through the ADHD lens, I decided to asked Claude (who “knows” I have ADHD) what might be happening. I described my experiences. Claude explained how these matched descriptions of ADHD shutdown. "He" responded with something that felt a lot to like compassion, understanding and kindness. As I recovered, I told Claude I was bored of lying in the dark and “he” advised I consider one of several low-cognitive drain recreational options if I had the capacity to choose. I did not. So Claude offered to “choose” for me. He discouraged me from rushing back to my daily life and I was pleased to discover, after taking Claude's advice, that I did not experience further smaller shutdowns, which had hitherto been the case.
I also discovered that Claude remembered what I told “him”. Two weeks later I said “I feel like this”, Claude flagged that these might also be part of shutdown and added these symptoms to the history. "He"also reminded me to tell my prescriber about them at our next appointment.
With humility, I must now concede that although Claude will give similar responses to the same question asked on separate chats, it isn’t necessarily “stock”. But would this experience encourage me to replace my very human psychotherapist with Claude? I'm not so sure.
To be clear, I’m willing to be open minded. I was not particularly in favour of online therapy pre-COVID and now it’s a core part of my practice. My ADHD shutdown experience with Claude has provoked me to think carefully about whether AI can be a useful mental health support. And so now feels like a good time to lay out my current thoughts on AI and psychotherapy.
So what do I think about AI as a mental health support?
The reader might be forgiven for thinking that I’m automatically going to say “no, bad idea” because of AI’s potential to disrupt my profession. But the reality is that the UK is facing a mental health crisis with the NHS and other auxiliary services overwhelmed by demand. So, a low-cost, easy access service that could help ease pressure and improve access to psychological services would be in everyone's best interest.
And my personal experience evidenced potential benefits. Claude did a good job of guiding me through the various stages of ADHD shutdown. So, I can see the benefit of AI in some instances. But we need to be realistic about what Clause was actually doing in that discussion and what AI can do overall.
Breaking down the support Claude gave me
ADHD shutdown is a physiological, neurological episode, not a psychological condition. That’s not to say that people with ADHD don’t face concurrent psychological challenges (which I will address in a blog post later next month) but at its core, ADHD is a biological difference in brain structure and function. Asking Claude to help manage recovery from ADHD shutdown is analogous to a diabetic asking Claude to help manage recovery after a hypoglyaemic episode (once medication has been administered as guided by an appropriately qualified physician). It does not map so easily to recovery from anxiety and depression. The former two conditions are about temporary shifts in body chemistry. The latter two are sustained, chronic adaptations in how a person's psychology filters and processes the world, sometimes exacerbated by changes in brain chemistry.
But it's only fair to let AI consider its own position on this too.
I asked Google’s Gemini what it thought about the comparisons made between ADHD shutdown, diabetes, anxiety and depression. It gave a similar interpretation to my own analysis when comparing the conditions.
Next I asked which would be better suited to AI support – ADHD shutdown or anxiety. It initially suggested AI factual support for ADHD shutdown (basically what I received) and what it called “a conversational AI intervention” for anxiety and depression, particularly to act as a cognitive sounding board. This supports the distinction already made about the nature of the two conditions.
I took this one step further and asked Gemini whether AI would be better than a human psychotherapist or counsellor. Gemini noted that conversational AI could help manage mild symptoms but lacked the capacity to build a true therapeutic alliance, which it cited as “the genuine human bond that is the largest predictor of successful therapy outcomes”. (Source: Sebastian Acevedo, Esha Aneja et al.).
Gemini also highlighted the risk that AI could reinforce harmful thoughts or give “wrong” advice. That’s not to say that all human therapists are perfect – in the UK, psychotherapist and counsellor are not protected titles, meaning anyone can call themselves that. For this reason, it is important to check a therapists credentials before working with them. But I was interested that Gemini highlighted there were concerns.
The "Articificial" nature of AI
Like Gemini, I have concerns about the limitations of AI.
AI is not human. It does not feel. It does not genuinely emote. So I challenge whether it can support transformational healing in the same way as human-to-human contact. Can a human being and an AI bot have what the Boston Change Process Group call “moments of meeting”, being spontaneous, intimate moments of deep connection between therapist and client where emotional barriers dissolve and authentic humanity is shared. These are the deep healing experiences that Gemini referred to earlier as the “genuine human bond”.
Music has suffered similar mimickery by AI. In 2023, a fan sent the singer-songwriter Nick Cave lyrics penned by ChatGPT in his style. He called them “a grotesque mockery” and a “travesty”. He went on to say “ChatGPT has no inner being, it has been nowhere, it has endured nothing, it has not had the audacity to reach beyond its limitations, and hence it does not have the capacity for a shared transcendent experience, as it has no limitations from which to transcend”. While AI can respond with words that mimic kindness and compassion, it cannot empathise because it cannot truly feel. As Cave put it, “algorithms don’t feel. Data doesn’t suffer”.
The most profound experiences of therapeutic change that I’ve experienced in my career have been punctuated by a shared understanding of pain and suffering. It’s the story of the person who falls into the hole as so beautifully told by Leo McGarry in the West Wing episode, Noel:
The guy who falls down a hole
This guy’s walking down the street when he falls in a hole. The walls are so steep he can’t get out. A doctor passes by and the guy shouts up, “Hey you! Can you help me out?” The doctor writes a prescription, throws it down in the hole and moves on. Then a priest comes along and the guy shouts up “Father, I’m down in the hole, can you help me out?” The priest writes out a prayer, throws it down in the hole and moves on. Then a friend walks by. “Hey Joe, it’s me. Can you help me out?” And the friend jumps in the hole. Our guy says “are you stupid? Now we’re both down here.” The friends says, “yeah, but I’ve been down here before and I know the way out”. (Credit: Aaron Sorkin)
I think this might be the crux of my issue with AI. None of these tools, no matter how friendly or charming they are have actually been down the hole before.
And that’s not to say that AI, along with medical intervention (vital in many instances, including for my ADHD) and, for those so inclined, prayer, are not important interventions. But where psychotherapy differs is the willingness to be at the bottom of the hole, where the walls are steep and the light feels a long way away. To be what Alice Miller calls “the enlightened witness”. While we might not have been in the exact same hole, we have been in holes like it, whether personally or professionally.
Our governing body’s requirement for personal therapy in training means that those of us governed by the likes of the BACP and UKCP have had to “be in the hole” with a “therapist friend” helping us find the way out of our own dark places. In the psychotherapeutic process, the experience of human suffering is shared, mirrored, held and ultimately transformed. Our wounds are formed in human interaction and so can only truly be repaired by human-to-human contact.
Alongside whether AI provides long term, deep repair, we also need to ask some important existential and ethical questions.
"Who" am I actually talking to?
When I speak to my therapist (yes, I still have one), I know I am speaking just to her. When I ask my AI assistant a question, it’s on my machine. It’s transmitted across the internet to company data centres and its read by cloud software with computers logging the text to track system errors and usage limits. Companies use data to train future models. While you can turn off chat history or delete data, back up copies may remain on secure servers through the system. In short, I don’t know where my questions and musings go when I “talk” to AI.
And what are the ethical codes governing the conversation?
When I speak to my therapist, I know she has decades of experience working within an ethical framework, plumbed into the NHS with deep experience navigating the UK’s mental health system. While you can ask an AI assistant to consider these codes, there’s always the caveat that they are not healthcare professionals and errors can occur. No such disclaimers apply to an accredited, registered mental health professional.
How do they form their responses?
When you ask an AI assistant a question or give it a scenario to consider, the assistant scrapes data to create its answer from sources which might be as reputable as the National Institute of Clinical Excellence but they also might not. While most quality AI Agents filter out lower quality data sources like unvetted personal blog posts, click-bait content farms, and anonymous social media forums, you cannot know for certain because most AI systems do not reveal their raw data sources in real time.
When you speak to any UKCP accredited psychotherapist, you know they’ve had to meet specific educational and experiential critiera, undertaking exacting training, including continuing to meet prescribed requirments for ongoing continual professional developemnt. Many of us gaining direct patient experience in the psychiatric system as part of our psychotherapuetic training. In short, we been trained in and gained real world, live experience recognizing the early and small signs heralding significant distress as well as being able to quote the diagnostic criteria.
Who is responsible for the responses being given?
Finally, we need to be very clear about the legal limitations of AI. Unlike a UKCP (or similar governing body) registered psychotherapist or counsellor, a clinical psychologist or psychiatrist, AI agents are not clinically and ethically responsible for the quality of the care they give. Although AI agents are improving constantly, we cannot ignore historic cases of AI agents encouraging suicide and, in one tragic case, offering to help a teenager write a suicide notes to their parents (source: The Guardian, 27 Aug 2025, accessed for this blog on19 Aug 2026).
So, why see a psychotherapist when you can talk to Claude for free?
I can only give my personal experience.
Claude mimicked compassion and kindess...
I found Claude mimicked compassion and kindness plausibly and gave sound guidance when I was experiencing the physiological symptoms of ADHD shutdown. And I suspect with time and development of large language models and RLHF (Reinforcement Learning from Human Feedback), AI agents will learn to mimic many aspects of the experience of being in a human-to-human therapeutic alliance.
...but my psychotherapist is willing to join me in the dark depths of the hole....
But, I have found my own psychological repair and reframing has happened because my therapist is willing to join me in the dark when I “fall down the hole”. While AI could give me a map and explain what’s going on, my therapist has been down there before...
