5 AI Prompts That Might Tell You More About Yourself Than You Expect

AI is normally presented as a tool for writing emails, summarising meetings and creating documents. But if you have used an AI over time and shared genuine context about your goals, decisions, frustrations, successes and failures, it can also become a surprisingly useful thinking partner.

Not a therapist. Not a doctor. Not an oracle. But potentially very useful at identifying patterns in the information you have chosen to provide.

These prompts work best when you give the AI enough honest context to compare over time, and when you ask it to show its evidence and challenge its own assumptions. A brand-new chat with almost no context may generate a generic personality reading rather than anything useful.

Before you start: use AI carefully

This is structured self-reflection, not therapy or diagnosis. If you have significant mental-health concerns, distress, thoughts of harming yourself or someone else, or symptoms that are affecting daily life, discuss them with a qualified health professional or an appropriate urgent support service. Do not use an AI conversation as a substitute for professional care.

Think carefully before entering highly sensitive personal, financial, medical or confidential information into any AI service. Check the privacy, retention, training and data-control settings for the product you are using. You can anonymise people, remove identifying details and describe a situation in less detail while preserving the question you want to explore.

Prompt 1: Who am I pretending to be?

Based on everything I’ve told you about myself, who do you think I’m pretending to be versus who I actually am? Don’t flatter me. Identify contradictions between how I describe myself and how I actually behave. Give me the evidence for each conclusion and distinguish observations from speculation.

Everyone has a story about themselves. You may see yourself as decisive, easy-going, ambitious, helpful or independent, while your actual behaviour sometimes tells a more complicated story. This prompt asks the AI to compare those two versions rather than simply repeat your preferred self-description.

The important safeguard is “give me the evidence” and “distinguish observations from speculation”. If the AI cannot point to something you actually told it, its conclusion may be an invented interpretation. You can also ask it to list evidence that would contradict the interpretation.

Prompt 2: What are my deepest insecurities?

Based only on patterns in what I’ve actually told you, what insecurities appear to drive some of my decisions? Rank the strongest ones. For each, show the behavioural evidence, give an alternative explanation, and tell me what evidence would prove your interpretation wrong.

The purpose is not to persuade an AI to reveal a hidden diagnosis. It is to examine whether some decisions repeatedly seem to be shaped by fear of failure, rejection, loss of control, looking inexperienced or disappointing someone. Those are possibilities to test against your own experience, not labels to accept.

The final sentence matters most. AI can construct a convincing narrative from a small number of examples. Asking what would prove it wrong forces the answer to remain provisional and gives you a way to notice confirmation bias.

Prompt 3: What cycle am I stuck in?

What is the single recurring behavioural cycle in my life that is most likely to limit me? Describe the trigger, my typical response, the short-term reward, the long-term cost and where I could interrupt the cycle. Be specific and use examples from what I’ve told you.

This may be the most useful prompt in the article. Many people repeat variations of the same pattern in different areas of life: avoiding an uncomfortable conversation, feeling temporary relief, then dealing with a larger problem later; taking on too much, receiving praise for being reliable, then becoming exhausted; or waiting for certainty, missing an opportunity and using the result as evidence that change is risky.

A useful answer should identify the trigger, the short-term reward and the long-term cost. That makes the pattern practical rather than merely descriptive. You are looking for an interruption point small enough to test, not a dramatic promise to change your entire personality.

Prompt 4: What happens if I fix it?

Assume I successfully break that cycle now. Give me a realistic picture of what my life could look like in one year, five years and ten years. Cover health, relationships, work, money, time and quality of life. Then show me the alternative trajectory if I don’t change it. Don’t assume everything works perfectly.

Small patterns compound. A decision to protect an hour each week, have a difficult conversation earlier or stop accepting work that cannot be done well may look insignificant today, but it can change what is possible over several years. The reverse is also true: an apparently small avoidance habit can consume time and options if it continues.

Ask for a realistic picture, not a motivational fantasy. Life contains illness, economic changes, other people’s decisions and bad luck. The value of this prompt is comparing plausible directions and making the cost of inaction visible, while remembering that no AI can predict your future.

Prompt 5: What is the highest-leverage change?

Taking everything above into account, what is the single highest-leverage change I should make during the next 30 days? Give me no more than five measurable outcomes. Tell me what I should stop doing as well as what I should start doing. At the end of the 30 days, give me five yes/no questions that will tell me whether I’ve genuinely changed my behaviour rather than simply been busy.

Insight without behavioural change is just entertainment. This prompt turns a long reflection into a limited experiment. “No more than five measurable outcomes” helps prevent an ambitious list that is impossible to maintain, while the stop-doing question makes space for the change rather than adding another obligation.

The yes/no questions should be about what actually happened: Did I have the conversation? Did I protect the time? Did I complete the agreed action? Did I stop the old response at least once? They should not measure how inspired or busy you felt.

Bonus prompt: make the AI criticise itself

Review your previous answers critically. Where might you have overinterpreted my behaviour, reinforced my existing beliefs, flattered me, or constructed a compelling story without enough evidence? What are the three conclusions I should be most sceptical about?

This may be the most important prompt of all. An AI is often rewarded for producing a coherent and satisfying answer. Coherence is not proof. Asking it to review its own reasoning can expose leaps, missing evidence, flattering language and conclusions that merely fit the story you already wanted to tell.

You should still do the final sceptical review yourself. Notice whether the answer relies on a single anecdote, treats a temporary situation as a permanent trait, ignores alternative explanations or sounds certain about information you never supplied.

How to get better answers

• Give concrete examples rather than asking the AI to infer everything from a mood or label.
• Ask it to separate observations, interpretations and speculation.
• Request an alternative explanation for every important conclusion.
• Ask what evidence would change the conclusion.
• Correct the AI when it misremembers or overstates something.
• Repeat the exercise after a meaningful period rather than asking the same question every day.
• Keep the final decision and responsibility with you.

An AI can help you organise a journal, compare decisions and ask questions you have been avoiding. It cannot observe your life independently, verify your memories, know what other people really think or guarantee that an interpretation is psychologically sound.

The IT Club view

The objective is not getting AI to understand you. It is using AI to help you understand yourself.

That distinction protects you from treating a fluent answer as an authority. Use the prompts to generate questions, find patterns worth checking and design a small behavioural experiment. Keep what is useful, reject what is unsupported and seek human or professional help when the issue is beyond a technology conversation.

Related reading

Is AI Saving Your Staff Time — or Creating More Admin? →

AI Employee Coaching: Useful Roleplay or Risky Replacement? →

How to Reduce AI Hallucinations in Business →

What Should You Know Before Putting Sensitive Information into ChatGPT? →

AI Agent Accountability: Who Is Responsible? →