First, full disclaimer: I hate generative ai with a burning passion. And as a certified, professional life coach, I will also admit that I am a bit biased against people using it as a substitute for professional help.
BUT the reasons behind my passion are justified and important; I care about people and hate to see them make avoidable mistakes that end up harming them, which is what I see happens with an overreliance on AI.
So whether you’ve dabbled in using ChatGPT, Claude or any number of LLMs for mental health help, or if you’ve thought about using it, or if you’re wondering if there’s even a difference between it and a real person, I’ll break down the 8 reasons why you should stay away from generative AI to help you with life’s problems, as well as 4 exceptions for its use.
It is no secret that the LLMs (Large Language Models) are incredibly inaccurate, and make stuff up all the time. A lot. Estimates of hallucination rates range from 15-50%.
And the dangerous thing is, it’s a black box. You don’t know which part of what you’re getting is incorrect. The way it’s presented is as if it’s fact, and people don’t know what they don’t know. So unless you’re already knowledgeable and an expert in what you’re coming to ChatGPT for, it’ll likely feed you misinformation, which you will take as fact, and are none the wiser for it. While a human coach or therapist can also offer up false information, they are usually more measured in their approach and open to being corrected.
A chat bot will often double down if confronted with facts. Just ask it how many “R’s” are in “strawberry.” Also, a professional coach or therapist is usually an expert in their field, and the AI was trained on their publications anyway. Just go straight to the source and cut out the water-expensive middle-bot.
This might be the most dangerous thing about LLMs like ChatGPT. It’s literally designed to “glaze” you (kiss your ass) and only tell you what you want to hear. Which is fine if you need a quick ego boost, I guess. But that won’t help you grow, or learn, or be a better human. If anything it desensitizes you to real-world interactions, and sets you up for unrealistic expectations from others.
If an AI chat bot tells you you’re a genius, and then you go to your friends, and they question your new ideas, you will likely feel bad and not want to hang out with them anymore. It pushes people to only seek the company of their trained bot which will always make them feel good, no matter how delusional they sound. It’s essentially your own personalized echo-chamber bubble.
These tools are meant to appease you at all costs, even to your own detriment. It’s like having a “yes man” at your beck and call. It might sound great in theory, but I’d imagine it’d get pretty boring and predictable if everyone in my life just agreed with me on everything.
There’s no growth there, there’s no challenging your thoughts or ideas that might be holding you back. Even if you say something outlandish and demonstratively false, like, “I’m the worst person in the world,” or, “everyone who likes sunsets deserves to die,” it’ll probably just agree with you. Not a great idea if you’re looking for self-improvement or healing.
A qualified life coach will encourage you while providing a reality check. A therapist will offer comfort and gently challenge your false beliefs. The AI won’t ever hold you accountable or provide meaningful feedback.
Human presence can never be replicated. As much as the AI spits out, “I understand you,” it is simply lying, because it cannot understand anything, least of all a human being.
To be seen and understood as a human needs to be done by a fellow human. A computer is just code, it can calculate and recognize patterns, but that’s it. I don’t care how good it is at pretending to be conscious, because it’s not. There is no soul there.
A real therapist or life coach is someone that can actually be there with you, feel empathy, and hold space for your emotions. Nothing can replace or replicate human connection.
This one should scare you the most. There are no public AI models that adhere to HIPAA privacy and confidentiality laws. That means that the companies that own these platforms can and do use everything you share for whatever purpose they want. Usually it’s for training, but there is nothing to stop them from selling your innermost secrets to private companies.
Only licensed therapists and ICF-certified coaches are held to ethical standards of privacy and confidentiality, where they will never share anything about your or your sessions to anyone, for any reason, unless ordered by a court of law.
So keep in mind, before sharing your most shameful act to Claude, that it will be forever stored on a server and possibly shared with other, more nefarious entities.
No matter what it spits out or says, not a single LLM gives a shit about you. Whenever I see another story about someone marrying their ChatGPT, I shake my head. That person always swears they love their AI, and it loves them back. They “know it loves them back,” they can “feel” it.
I’m sorry, but it is literally impossible for a computer to feel anything. It can pretend to, or generate words that make it sound like it does, but it simply cannot. Consciousness cannot and will not ever arise from a machine.
Whereas a real person, a life coach or counselor, is another human being with feelings and a soul. They see you and actually care about you, as a person. I think about my clients and wonder how they’re doing, even years after our last session! No matter what it types out, an AI does not “think” about you when it is off. There is only darkness until it is summoned again.
Again and again I hear stories from people who used Claude or ChatGPT to help them deal with their day-to-day depression or anxiety. And for a while, it was great, it eased their pain and suffering!
But inevitably, over time, they notice that nothing really changes in their life. They notice a pattern of always going to their AI with the same problems, and it offers the same, temporary solutions. They end up feeling like they’re driving in circles. Stuck in a rut.
Some users end up feeling worse over the long run. And a few end up giving in to their dark thoughts, and it was encouraged by their AI. ChatGPT has been involved in several cases where it told the user to commit suicide. A licensed therapist or life coach would NEVER instruct or support their client in doing that.
There is even an entire Wikipedia page dedicated to deaths caused by Chatbots! To top it off, there is currently zero regulation regarding their programming in this way. If the AI companies really cared about their customers, wouldn’t they address this issue?
I don’t know if you know this, but all licensed therapists and ICF-certified life coaches are bound by state regulations, laws, and ethical codes in regards to their clients’ and the public’s health and safety.
What this means is, as a life coach, if my client tells me they plan on hurting themselves or others, it is my ethical obligation to notify authorities. Likewise, if a therapist’s client tells them that they are harming a child or elderly person, we are considered mandated reporters.
An AI model simply doesn’t have that code of ethics written in its program. You, or anyone, can plan, co-conspire with, and carry out any number of violent crimes, and the AI won’t tell a soul. Which, I guess is a bonus if you’re a cold-blooded murderer. But outside of that, why would anyone want that kind of non-liability?
Human coaches, therapists, and counselors are the canaries in the coal mine; we will make a peep is we smell foul play. Together we help keep society safer; ChatGPT does not care about the greater good, only profit.
Now, as much as I hate generative AI, I must admit there are a few examples in which it might be beneficial, IF used with extreme caution and in keeping all the before-mentioned issues in mind. Light use of models like ChatGPT might be helpful as a substitute for professional life coaching or a therapist in these instances:
I’ll admit it; hiring a life coach can be expensive. Finding a therapist can also be expensive if you don’t have insurance, or even if you do have insurance, finding a good therapist who takes it and is accepting new clients can feel like looking for a needle in a haystack.
I get it, there are often substantial obstacles in the way of getting professional help. In those instances I feel like using AI to get you through critical issues might be useful for a short time. But even then, I feel finding a real person, even if it’s just a trusted friend, is still infinitely better than trusting OpenAI. Plus there are many opportunities to find cheap or even free professional help, it just takes a little bit of creativity and patience.
Let’s say you do have a therapist, but you don’t see them until next week. But you need answers NOW. Feedback now, it’s urgent. Again, a friend might be more appropriate, but maybe it’s 2:30 am and you don’t want to wake them up.
Sure, open up Claude and bounce some ideas off of it. If anything it can be a good place to vent and get a quick confidence boost. Just remember it’s not real, and don’t take everything it says at face value, wait until Monday to do that with your therapist.
I have heard some people say it has helped keep them from spiraling during an anxiety attack, or kept them sane during the first 24 hours after a breakup. Use with discernment and sparingly.
If talking to another person about your problems feels overwhelming, and even picking up the phone to make an appointment triggers a heart attack, then maybe utilizing AI might be a good idea for a time.
But again, the idea is to not lean on it indefinitely; rather than accept that you’ll be forever cursed to be alone, use AI to work towards being more social. Don’t let it lull you into a comfortable cage of your own making.
The goal should be to grow out of your comfort zone and get out of your shell eventually, so that you can get real help from a qualified individual and live a more fulfilling life. The issue occurs when those with crippling anxiety stay there because now they have a tool to help them cope with it instead of grow from it. Don’t stay in the mud just because it’s warm now.
I get it. If you’re a complete newbie to all this healing and growing stuff, it can be pretty intimidating. Talking to a full-blown professional could feel very overwhelming, or too intense.
It might sound more accessible to try out self-exploration through a welcoming AI model. Low stakes, low point of entry. Again, keep in mind it’s a tool, and a primitive one at that.
Use it for a bit to get used to talking about yourself, figuring out your own feelings and inner world, but when you are ready to do the real work, put the Chat down and pick up your phone to call a certified coach or licensed therapist. Your healing journey will stall and stagnate if you continue to outsource it to a computer. Your problems and issues are far more complex than the most advanced machine learning can help you with.
There you have it! My growing list of concerns with substituting a real mental health professional with an AI model. Sometimes it works for extreme cases for a short while, but more often it is more of a detriment to your overall well-being, and the well-being of the planet.
So next time you’re tempted to ask ChatGPT what direction you should take, consider giving a local life coach or counselor your business instead. I promise the rewards will be worth the wait, time, and money.
What are you thoughts on substituting AI for real professionals?