The short version: ChatGPT can help you spot patterns in how you think, decide and react by asking you the right follow-up questions, but it cannot test your personality the way a real psychometric tool does, so use it as a mirror for reflection, not a diagnosis. The prompts that work best ask for evidence from your own life, not vague adjectives. I have run this exercise on myself for three months and it changed how I schedule my week more than any personality quiz ever did.
Why I started doing this at 53
Five years ago my business fell apart quietly, the way these things do. Not one dramatic event, just contracts ending, an influencer career that had run its course, and me sitting at my kitchen table in Berkshire trying to work out what I wanted to rebuild rather than just what I was good at faking. I had done the Myers Briggs test three times over twenty years and got a different letter combination each time. I did not trust it anymore.
What I trusted was catching myself mid-decision and asking why I made the choice I made. ChatGPT turned out to be a decent tool for that, not because it knows me, but because it asks a follow-up question when a human friend would just nod and change the subject. I started typing things into it late at night, half journalling, half interrogating myself. Within a few weeks a pattern showed up that I had never named out loud: I say yes to client calls I do not want to take because saying no feels like abandoning someone, even a stranger. That single insight, surfaced through a chat with an AI model, changed how I built my calendar for 2026.
What ChatGPT can and cannot tell you about yourself
Here is the uncomfortable bit nobody selling you a "personality prompt pack" wants to say plainly. ChatGPT is not assessing you. It has no access to your behaviour, your relationships, or a control group of ten thousand other people to compare you against. It is doing something much simpler: taking the words you type and reflecting them back with structure and a confident tone. If you tell it you are anxious and perfectionistic, it will find evidence in your own sentences for anxiety and perfectionism, because you handed it the raw material.
This is not a flaw you can prompt your way around. It is how the technology works. The Cambridge Dictionary defines a personality trait as an enduring pattern of thought and behaviour, and enduring patterns get proven by what you do across months and years, not by what you type into a chat box on a Tuesday night. So the value here is not "ChatGPT told me I am an INFJ." The value is that answering its questions forces you to notice your own patterns out loud, which most people never do because nobody asks them the second question.
Real personality assessment, the kind psychologists use, comes from validated instruments like the Big Five model (openness, conscientiousness, extraversion, agreeableness, neuroticism), which has decades of research behind it. You can read a proper overview of that model on Wikipedia's Big Five personality traits page if you want the academic version. Use ChatGPT alongside that framework, not instead of it.
The exact prompts I use, in order
I run this as a fifteen to twenty minute session, usually on a Sunday evening, not scrolling on my phone but sat at a laptop with no other tabs open. Here is the sequence I type, word for word:
- "Ask me one question at a time about a recent decision I struggled with, and after each answer, ask a follow-up before moving on."
- "Here are three moments from the last month where I reacted more strongly than the situation seemed to need. [Describe them in two or three sentences each.] What pattern do you notice across all three?"
- "Based on what I've told you, what am I avoiding? Be direct, don't soften it."
- "If a close friend described my behaviour in these examples without using my name, what trait would they name?"
- "What is one small experiment I could try this week to test whether that trait is true, rather than just a story I tell about myself?"
Notice what these have in common. None of them ask ChatGPT to label you from a list of sixteen types. All of them ask it to work from specific examples you supply. That is the difference between a prompt that produces something useful and one that produces flattering nonsense that reads well and means nothing.
A real session, and what came out of it
In September I gave it this: "Three weeks ago a client pushed our call from Tuesday to Friday with two hours' notice. I said 'of course, no problem' and then felt resentful for two days. Last week my daughter asked me to help with a project the night before it was due and I stayed up until 1am doing it, again saying it was fine. Yesterday a supplier chased an invoice I owed and I paid it late because I had quietly deprioritised it, even though the money was there."
I asked what pattern connected the three. The answer, paraphrased: in the first two examples I gave my time away instantly and paid emotionally for it afterwards, in the third I withheld something small (a prompt payment) from someone with less power over me than the client or my daughter. It suggested the trait underneath might not be "people pleasing" in general, but something narrower: I protect relationships where I fear disapproval and quietly punish or delay in relationships where I don't. That is not a Myers Briggs letter. It is a specific, checkable claim about my behaviour, and when I looked at the past two months of invoices, it held up. I had paid four suppliers late in eight weeks, all small businesses, none of them clients I was afraid of losing.
That single conversation cost me nothing and took eleven minutes, and it told me more about a real pattern than any twenty-question quiz I had taken in the previous decade. The quiz gives you a label. The conversation gives you a data point you can test.
Step by step: running your own session
If you want to try this rather than just poking at ChatGPT for five minutes and giving up, here is the structure that has worked for me and for a handful of clients I have walked through it:
- Step 1: Pick one week, not your whole life. Recent memory is more honest than a general self-image.
- Step 2: Write down three specific moments, with what happened, what you felt, and what you did. Two sentences each, no more.
- Step 3: Ask ChatGPT to find the pattern across those three moments only, not to guess at your whole character.
- Step 4: Push back at least once. Type "give me the opposite explanation" and see if the second answer is more honest than the first.
- Step 5: Ask for one small, testable action, then do it within the week and check back in with what happened.
- Step 6: Repeat monthly. One session tells you a story. Six sessions across six months tell you whether the story is true.
Step 4 matters more than people think. A single AI answer, delivered confidently, feels like truth. It is not. It is the first plausible pattern the model found in your text. Asking for the opposite explanation forces a second draft, and often the second draft is the sharper one.
Prompts for specific traits, if you want a starting list
Some readers want a menu rather than a method, so here is one, built around traits people ask me about:
- Perfectionism: "Describe three tasks I delayed finishing in the last month. What was I afraid would happen if I finished and shared them?"
- Conflict avoidance: "Here is a disagreement I had at work where I backed down. What did I gain by backing down, and what did it cost me?"
- Impulsivity: "List three purchases or commitments I made without sleeping on them. What emotion came right before each one?"
- Need for control: "Here's a situation where I struggled to delegate. What specifically did I fear about letting someone else do it?"
- Self-worth tied to output: "On days I don't finish my to-do list, how do I talk to myself, based on these examples? Would I say that to a friend?"
Each of these forces evidence before opinion. That is the entire trick. Any prompt that starts with "what personality type am I" is asking the model to guess. Any prompt that starts with "here are three specific things I did" is asking it to analyse, and analysis of real examples is where the useful insight lives.
Want AI doing the heavy lifting in your marketing?
I build the systems that handle the boring 80 percent, so you get your week back. Done properly, with the human kept in.
Where this goes wrong
I have watched people, including a version of myself in 2021, use this as a substitute for actual reflection rather than a support for it. You type a paragraph, get a warm, articulate answer back, feel understood for four minutes, and move on without changing anything. That is not self-knowledge, that is a nicer version of scrolling.
The other failure mode is treating a chatbot's read on you as more objective than a person who knows you, purely because it sounds calm and structured. It has no memory of who you were five years ago, no idea how you behave when you are exhausted or grieving or in love, and no stake in whether the insight helps you or just sounds clever. Run the pattern past one real person who has known you for years before you build a life decision on it. I check my own AI-generated insights against my husband and against my business partner before I act on them, and about a third of the time they tell me the pattern is real but I have the cause wrong.
If you are doing this alongside real work on your business or your career, and you want a structured look at how AI tools fit into that bigger picture rather than just personal reflection, that is a different conversation, and one worth having with someone who has done it for other founders rather than working it out alone at midnight.
This guide is part of my ChatGPT Prompts and Limits: 30 Guides to Fun Prompts, Photo Edits and Daily Caps.
If that raises the next question, see What Are Some Fun ChatGPT Prompts for Personal Use? 47 I’ve Tried.
Try my free prompt builder before you post.
If you'd rather chat with a ready-made AI character than write the prompts yourself, my guide to Character AI alternatives covers the safer options.
I've also explained persona design for virtual assistants.
Frequently asked questions
Can ChatGPT tell me my Myers Briggs or Big Five type?
Not reliably. It can hold a conversation shaped like one of those tests, but it has no validated scoring system behind it, so the result reflects your wording that day more than a stable trait. Use it to explore patterns, not to generate an official type.
How often should I do this kind of self-exploration with ChatGPT?
Once a month works well because it is close enough to notice change but far enough apart that you have fresh examples to draw from. Weekly sessions tend to just repeat the same three anecdotes and stop producing new insight after about a month.
Is it safe to share personal stories with ChatGPT for this kind of exercise?
Avoid putting in identifying details about other people, like full names or addresses, and remember your conversation may be used to improve the model unless you have adjusted your data settings. Keep it to your own reflections and change small details if you are unsure.
What is the biggest mistake people make when using AI prompts for self-reflection?
Asking for a label instead of an explanation. "What's my personality type" gets you a guess dressed up as insight. "Here are three specific things I did this week, what pattern connects them" gets you something you can test against your own behaviour.
