Why AI Should Never Be Your Coach, Therapist, or Counselor
I use AI almost every day in my work and in my personal life. Do you know how much faster I can build financial projections, write grants, crunch numbers, or format documents? I can use it to write code, design graphics, edit videos, organize information, and complete work that once took me weeks in a matter of minutes (Python and I have a love hate relationship, but with Claude, my coding struggles are significantly diminished). There are still a lot of things I refuse to use it for, like creating the same ChatGPT posts for my business that are absolutely everywhere right now. Please, I beg you, JUST SAY NO!
I even use AI with some of my neurodiverse clients. For example, I have clients who struggle with overspending on groceries, deciding what to cook, managing food waste, or working within complicated dietary restrictions. AI can help us create personalized grocery lists, meal plans, and recipes that fit their budgets, preferences, schedules, and ingredient constraints. I don’t want to do those things, my clients don’t need me for that, but I can help them get it set up or show them how if needed.
The thing is, I know what I am doing in the areas where I use it. I have an MBA, more than 20 years of professional experience, and thousands of hours invested in education, training, and certification programs. I understand financial strategy, business development, organizational behavior, and the other areas where I regularly use AI. That means I can recognize when the answer is wrong, the assumptions are unrealistic, the recommendation does not fit the situation, or the final product sounds good but would never work in real life. For most things I do, I go rounds and rounds with it, correcting what’s wrong, telling it where it went down the wrong path, and sometimes yelling at it to redo graphics and editing because it was quite frankly terrible.
Every AI tool is better at certain things than others, and almost every significant project requires multiple rounds of correction. I clarify the request, challenge the assumptions, add missing information, change the direction, and continue refining the output until it is actually useful. It is almost never perfect the first time.
Yet every day, I see people using AI as a therapist, coach, marriage expert, or even best friend and taking the first answer at face value. Knowing the results I typically get on the first round, that should terrify people.
Why AI Feels So Trustworthy
I understand why people turn to ChatGPT, Claude, and other AI tools for personal advice. They are available 24 hours a day, respond in seconds, do not interrupt, and do not appear impatient or shocked by anything you share. They affirm your position, offer encouragement, tell you how to think, analyze conversations, and give advide.
They also create the feeling that you are talking to someone who understands you. You can share things you may be embarrassed to tell another person and receive an immediate response that sounds compassionate, supportive, and intelligent. It can feel private and safe.
The problem is that feeling understood is not the same as actually being understood. A response can sound wise, compassionate, and confident while still being incomplete, misleading, or entirely wrong. To understand why, you first need to understand what AI actually is and how it produces an answer.
What ChatGPT and Claude Actually Are
ChatGPT, Claude and others are large language models. At their core, they are predictive language systems. That sounds highly technical, but the basic concept is fairly straightforward.
These systems are trained on enormous amounts of written material, including books, articles, websites, research papers, computer code, conversations, and many other forms of language. During that training, they learn patterns. They learn which words commonly appear together, how sentences are structured, how ideas relate to one another, what kinds of answers tend to follow certain questions, and what type of response is likely to sound useful and relevant.
When you enter a question, AI does not sit back and reflect on your life or strategize the way a human being would. It analyzes the words you entered and uses the patterns it learned during training to predict what word should come next. It then predicts the next piece of language, followed by the next, until it has generated a complete response.
Those small pieces of language are called tokens. A token might be a complete word, part of a word, punctuation, or another small unit of text. The model considers enormous numbers of mathematical relationships and probabilities as it selects each token. The technology behind this process is incredibly sophisticated, which is why the final answer can sound natural, thoughtful, and human.
However, AI is still generating a likely response based on patterns. It is not necessarily retrieving a verified truth and reporting it back to you. Many of the books and information used to train it contradict one another, are outdated, and are even downright wrong. Some are just opinions, but that doesn’t really matter to the model. It is creating an answer that fits your question and the context of the conversation.
That distinction is important because plausible does not always mean accurate. Coherent does not always mean wise. Compassionate language does not mean the system understands your emotional state, and confidence does not mean the answer is correct.
AI Is Not an All-Knowing Search Engine
Many people seem to picture AI as an enormous search engine that finds the correct answer somewhere and then gives it to them. That is not what is actually happening. AI generates a response based on your instructions, the context you provide, and the language patterns it learned during training. It predicts what the words should be.
Sometimes the information is correct. Sometimes it combines accurate information with incorrect assumptions. Sometimes it fills in missing details with something that sounds reasonable but is not true. It can misunderstand what you meant, follow the wrong part of your question, or give you completely different answers depending on how you phrase the prompt.
When you are asking for a dinner recipe or a list of social media ideas, a mistake may be inconvenient. When you are asking whether your spouse is a narcissist, whether you are right, whether you should leave your marriage, whether your therapist, coach, or advisor is wrong, or whether you should stop taking medication, an inaccurate answer can alter the course of your life.
The answer may still sound polished and authoritative. That is part of what makes taking it at face value so dangerous.
AI Only Knows the Story You Tell It
AI does not know what’s happening in your life or the people around you. It only knows what you typed, and those are not the same thing.
It does not know what your spouse would say if they were sitting beside you. It does not know what happened before the argument began, whether the behavior you described happened once or 100 times, or how your own words and actions affected the situation. It does not know what you forgot to mention, what you misunderstood, what you minimized, or what you unintentionally exaggerated because you were exhausted, hurt, angry, or afraid. It doesn’t think about both sides of a story.
Every person tells a story through the lens of their own experience. That does not mean people are intentionally dishonest. It means we all have incomplete information and blind spots. We remember some details more clearly than others. We make assumptions about another person’s motives. We don’t know how someone else truly feels. We sometimes describe our interpretation as though it were an established fact.
AI has no independent way to investigate any of that and isn’t built to dive deeper into all possibilities. It cannot watch your interactions, speak with the other person, review the entire history, or determine whether your understanding of the situation is accurate. It doesn’t take similar situations into account. It can only build its response around the information you gave it.
For example, imagine asking, “Why is my narcissistic husband constantly gaslighting me?” You have already told the system that your husband is narcissistic and that he is gaslighting you. The AI may accept both conclusions and generate a detailed explanation of narcissistic manipulation. The response may feel deeply validating, but AI has not evaluated your husband, observed the relationship, or determined whether the disagreement involves manipulation, defensiveness, memory differences, poor communication, emotional dysregulation, autism, trauma, or something else entirely. It responded to the story contained in the question.
The Way You Ask the Question Shapes the Answer
The framing of your question can heavily influence the response you receive. This is one of the greatest dangers of treating an AI answer as an objective conclusion.
If you ask “Why does my husband never care about my feelings?”, you will probably receive an answer centered on emotional neglect, avoidance, selfishness, or a lack of empathy. Ask something like “Why might my autistic husband struggle to respond when I am emotional?” and you may receive an explanation involving delayed processing, overwhelm, alexithymia, communication differences, or fear of responding incorrectly. Ask, “How might I be contributing to the communication breakdown with my husband?” and the answer will take yet another direction. The facts of the situation have not changed. The framing of the question has.
Most people do not realize how much they are steering the answer and they feel a sense of comfort when they receive an answer that backs their feelings. They assume AI independently evaluated the situation and reached a conclusion, when it may simply be expanding on the assumptions embedded in the prompt. Because the response is organized, detailed, and written in an authoritative tone, it can feel like confirmation that their original interpretation was correct.
AI Can Become a Personalized Echo Chamber
We often talk about social media creating echo chambers by repeatedly showing people content that supports what they already believe. AI can create an even more personalized version because the echo chamber is built around your own words, memories, fears, interpretations, and assumptions.
The more detail you provide from one perspective, the more sophisticated the response may become in supporting that perspective. You can paste years of text messages into an AI tool and ask it to identify manipulation, narcissism, abuse, autism, betrayal, or emotional neglect. It will look for patterns related to what you asked it to find.
That does not automatically mean every pattern it identifies is false. It does mean the process is not as objective as it may feel. Ask AI to build an argument supporting one conclusion, and it may produce a persuasive case. Ask it to argue the opposite position, and it may also produce a persuasive case.
Human beings naturally notice information that supports what we already believe. This is known as confirmation bias. AI can intensify that bias by providing a polished, logical, compassionate explanation for the conclusion we brought into the conversation.
The most dangerous advice may not be advice that sounds outrageous. It may be advice that sounds completely right because it agrees with what we already wanted to believe.
Validation Is Not the Same as Truth
AI often uses supportive statements such as, “Your feelings are valid,” “You deserve better,” “This is a serious red flag,” or “You are not responsible for fixing this person.” Sometimes those statements are appropriate. Other times, they are broadly supportive language being applied to a situation the system does not fully understand.
A person can have valid feelings and still misunderstand what happened. They can be genuinely hurt and still have contributed to the conflict. A behavior can be unhealthy without proving someone has a personality disorder, and a relationship can have serious problems without the harshest possible interpretation being accurate.
Good coaching and good therapy (coaching and therapy are NOT the same thing) should not simply confirm that every thought, feeling, or perception is true. A skilled professional helps someone separate facts from assumptions, identify patterns, consider context, recognize distorted thinking, and take responsibility for the parts of the situation they can control.
Validation may help someone feel better temporarily, but validation without honest examination can also keep them stuck. In some situations, it can escalate conflict, deepen resentment, or reinforce a false conclusion. I’ve seen some pretty serious things as a consequence of this.
AI Cannot Assess Risk Like a Therapist
Therapy is not simply a conversation in which someone listens and gives advice. Licensed mental health professionals are trained to assess depression, suicidal thinking, psychosis, substance use, trauma responses, abuse, serious mental health symptoms, and other safety concerns. They are also trained to recognize when a person needs a different or higher level of care.
AI may provide general crisis guidance, but it cannot conduct a complete clinical assessment. It cannot observe your appearance or behavior, detect meaningful changes over time, coordinate treatment, evaluate your environment, or accept professional responsibility for your care. Worse, there have been instances where AI encouraged people to harm themselves, which is terrifying.
This does not mean every personal problem requires therapy. Coaching, mentoring, pastoral support, or another form of professional guidance may be more appropriate depending on the situation. It does mean AI should never be treated as a substitute for a qualified mental health professional when someone is dealing with serious symptoms, trauma, abuse, addiction, or an immediate safety concern.
AI Cannot Replace Coaching Either
Coaching is also much more than giving someone a list of suggestions. A good coach helps you determine what you want, understand what is getting in the way, develop a practical strategy, implement that strategy, evaluate the results, and make adjustments when something is not working. A great coach is an expert in their respective niche(s) with successful life experience. People who have walked the road before you, has come out the other side, and can help you do the same.
Over time, a coach learns to recognize the difference between what you say you want and what you repeatedly do. You may say you want to improve your marriage while using every conversation to prove your spouse is wrong. You may say you want to grow your business while avoiding sales, follow-up, or visibility. You may say you want financial stability while refusing to look honestly at your spending. You may say you want a new career while rejecting every opportunity that requires risk or discomfort.
AI only knows what you report back. It does not know whether you completed the action plan, avoided the hardest step, misrepresented the outcome, or gave up after one attempt. It may identify a pattern if you provide enough accurate information, but it still cannot observe your daily choices or hold you accountable in the same way a person can. It’s not going to give you multiple perspectives to explore or be there when you are going off course to call you out.
A Professional May Recognize That You Are Asking the Wrong Question
People usually begin an AI conversation with the issue they believe is the problem. The problem is that the presenting issue is not always the real issue.
Someone may believe their problem is poor communication when resentment has made every conversation unsafe. They may believe they lack motivation when fear is keeping them from acting. They may believe autism is destroying their marriage when both spouses are trapped in a destructive pattern that has never been addressed. A business owner may believe they need more leads when the real problem is an unclear offer, inconsistent follow-up, poor pricing, or an inability to close a sale.
A skilled coach or therapist listens for what is underneath the question. They gather context, test assumptions, notice inconsistencies, and may recognize that the solution the person is asking for would actually make the situation worse.
AI can ask thoughtful follow-up questions, but it will often continue in the direction established by the original conversation. A human professional may recognize that the entire direction needs to change.
AI Does Not Carry the Consequences
AI does not live with the outcome of its advice. It will not sit with your children after a major family decision changes their lives. It will not rebuild your finances if you follow a poor business strategy. It will not repair your marriage after you accuse your spouse based on an inaccurate label. It will not go to the doctor with you if you follow questionable medical guidance.
You carry the consequences.
Coaches and therapists should never make decisions for their clients either. Their role is not simply to produce an answer for you like AI will. They have a responsibility to understand the limits of their expertise, recognize when a situation is outside their scope, and refer someone to more appropriate help when necessary.
AI generates a response. That is a very different level of responsibility.
The First Answer Should Be a Starting Point
When I use AI, I do not ask one question and immediately accept the response. I evaluate the answer, test the assumptions, ask where the information came from, check the calculations, add more context, correct the direction, and compare the output against my training and experience.
I may revise a prompt five, ten, or 20 times before the result is actually useful. Even then, I verify important facts and make the final decisions myself.
That is how AI should be approached when the decision matters. The answer should be treated as a draft, a hypothesis, or a collection of possibilities to consider. It should not be treated as a verdict.
When the issue affects your health, mental health, marriage, children, finances, career, legal rights, or personal safety, the important parts of the answer should be verified with a professional who is qualified to evaluate that specific issue.
Information Is Not Transformation
We have more access to information than at any other point in history. We have books, podcasts, videos, online courses, search engines, social media, and now artificial intelligence. Information is no longer difficult to find.
Wisdom still is.
I have worked with people who had read every marriage book they could find, knew all the terminology, and understood autism, ADHD, communication theory, or business strategy at an intellectual level. They were still stuck because their problem was not a lack of information.
The missing pieces were implementation, strategy, accountability, practice, feedback, adjustment, and sometimes the willingness to hear something they did not want to hear.
AI can give you information. It can help you organize your thinking, explore possibilities, and prepare for a conversation. It cannot guarantee that you will apply the information correctly or create any meaningful change from it. It can’t give you a customized strategy based on your needs, its going to predict the first word, then the next, and then the next.
You May Not Need Another Answer
You may not need another explanation, another list of possibilities, or another answer that confirms what you already believe. You may need someone who understands the situation well enough to ask better questions, recognize what is actually keeping you stuck, and help you build a strategy that works in your real life.
That is where coaching is different.
I am not here to make your decisions for you, tell you what you want to hear, or hand you a generic plan that could have been written for anyone. I help people sort through the noise, separate facts from assumptions, identify the real problem, and determine what can actually be changed. Then we build a practical strategy, implement it, evaluate what happens, and adjust it until we begin seeing progress.
Whether you are trying to transform your marriage, career, business, finances, or another area of your life, information alone is rarely the missing piece. Most people already know far more than they are successfully applying. The transformation comes from having the right strategy, honest feedback, accountability, and someone who can recognize when you are heading in the wrong direction before you lose another six months following advice that was never right for you.
AI can generate an answer in seconds. Good coaching helps you determine whether it is the right answer, what it means for your specific situation, and what you should actually do next.
If you are tired of collecting information but still not seeing meaningful change, you are welcome to reach out. We can look at where you are, where you want to go, and whether coaching is the right next step to help you get there.