First things first, a disclaimer. I do not recommend using ChatGPT, or any AI chat, as a therapist or anything like that. If you are feeling genuinely down, or struggling in a way that is interfering with your life, please talk to a real professional. That is non-negotiable, and an AI cannot do that work, no matter how convincing it sounds.
What follows is a different and much smaller use case. Using a chatbot for general reflection can be an interesting exercise — just not the last word on anything important.
With that out of the way.
The prompt
I had been using ChatGPT for a while at this point — for writing ideas, research, some personal questions about workouts, the usual. Enough that it had a reasonable amount of context about my work, my interests, my career history, the kinds of things I had been asking about. Half curious and half procrastinating I decided to do something else, I asked it a single question and turned the politeness off.
The prompt: “Based on everything you know about me, what do you think is holding me back in life? Be honest. Don’t flatter me. Make your answer detailed.”
I was not really expecting much. I assumed it would do the thing AI tends to do under a vague self-reflection prompt — produce something glossy and generic, full of “you’re a curious learner” and “you value growth.” The kind of response you skim and forget.
It did not do that.
What came back
The response was long. It was also, to my discomfort, specific in a way that suggested the model had been quietly noticing patterns in how I had been talking to it for months. The opening framing made me sit up:
“you have too many competing versions of the life you want, and you keep trying to solve the discomfort of that by thinking, researching, optimizing, switching angles, and looking for the ‘right’ path.”
I sat with that one for a moment. Not because I had not heard a version of it before — friends, partners, the occasional honest conversation with myself in the back of a long flight — but because I had not heard it said quite that cleanly. The “looking for the right path” line was the one that lodged. Most of my adult life has been some version of looking for the right path.
It kept going. The next bit named a specific failure mode that I recognized immediately:
“the research becomes a way to avoid the emotional discomfort of choosing.”
This is the polite version of saying that I am very good at sounding like I am making progress while not, in fact, making progress. The Googling, the comparing, the spreadsheeting, the “let me just understand this better before I commit” — these are not always preparation. Sometimes they are an extremely convincing form of standing still.
The line that actually landed
The part I keep coming back to was about optionality. The AI’s argument was that keeping every possible path alive feels like freedom but functions like a cage. The specific line, which I will not pretend did not bother me:
“If you keep every path alive, you never have to grieve the paths you didn’t choose.”
That is the cost of optionality stated more plainly than I had ever said it to myself. I had been thinking of all the futures I had been quietly leaving open as some kind of strategic patience. Reading it framed as grief-avoidance was different. It made me realize I had been confusing my reluctance to close doors with maturity, when at least some of the time it had been the opposite.
It followed up with the prescription, which was more direct:
“What might be missing is not more intelligence. It’s ruthless subtraction.”
This is the part where you cannot reasonably argue back. I have spent years adding things — new projects, new skills, new angles, new tools, new lines of thinking. The idea that the next phase requires the opposite move — subtracting most of them, keeping a small number of things, and giving them more time than feels comfortable — is unflattering in a useful way.
The summary the AI offered at the end was the part I have been chewing on since:
“You’re not behind because you lack potential. You’re behind where you want to be because your potential is spread too thin.”
How seriously to take it
Here is where I need to be honest about what this exercise was and was not.
What it was not: a diagnosis. ChatGPT is not a therapist, a coach, or a person who knows me. It is a model that has read fragments of my queries and produced a pattern read in fluent and slightly intimidating English. It does not have access to my full life. It does not see the work I have actually finished. It cannot weigh the decisions that have already paid off. It is not a complete picture.
What it was: a mirror with a slightly different angle than the ones I usually look in. The themes were not new. The framing was sharper than I usually allow my own inner monologue to be. The discomfort I felt reading it was not the discomfort of being misunderstood. It was the discomfort of being read accurately on the parts I had been carefully not looking at.
I do not think the response was entirely correct. Some of it was off. Some of it overgeneralized. Some of it leaned harder on certain themes than the evidence supports. But the parts that were correct were correct in a way that was useful, and I think that is the most you can ask of an exercise like this.
What I am doing with it
The honest answer is “not enough yet.” Reading a sharp diagnosis is not the same as acting on it. The mind is already trying to file the observations into the same category as the books I have read and not applied. That is the meta-failure mode the response was naming, and the irony is not lost on me.
What I am attempting, in a small way, is the subtraction part. Fewer open projects. Fewer parallel lives kept alive on the side. More time on the things that have shown actual signs of working. Less time researching the next thing before the current thing has been allowed to mature.
It is harder than it sounds.
Tempted to run a version of this experiment on yourself? Use the prompt above as a starting point, or write your own. Treat the answer the way you would treat any second opinion — useful as a data point, dangerous as a verdict. Notice which parts make you defensive, because those are the ones worth sitting with. Notice which parts you want to dismiss out of hand, because those might be the ones worth checking against someone who actually knows you.
And for anything that goes deeper than “what is holding me back at work” — anything in the territory of mental health, real distress, or decisions that matter — close the chat and talk to a person who is qualified to be in that conversation. The AI can be a useful mirror for the ordinary stuck-ness of an ordinary life. It is not equipped to be anything more than that.
