How to Use AI to Find Your Strengths: 8 Self-Assessment Prompts for Your Values and Best-Fit Work, and the Limits of the Answers

You are getting ready for a job search or a career change, taking stock before a performance review, or sitting in an interview when someone asks you to "sum up your greatest strength" and your mind goes blank. AI can be a genuinely useful partner at moments like these. But if you simply ask it "What kind of person am I?", all you get back is a plausible-sounding expansion of whatever you typed in.

What works is the reverse. Let the AI ask the questions to draw the material out of you, have it help sort out the common threads and put them into words, and make the final call yourself. This article gives you 8 prompt templates for that (strengths, values, weaknesses, the kind of work that suits you, a personal pitch, mock interview questions and more), along with why you shouldn't take the AI's answers at face value, how to handle personal information, and how to use ChatGPT, Claude and Gemini for the job.

1. What to hand to AI, and what not to

AI is good at asking questions, spotting common threads in scattered stories, and shaping what you want to say into clear sentences. On the other hand, if you ask an AI that doesn't know you to "judge" you or to "fill in facts", you get answers that sound convincing but have nothing behind them. Drawing the line up front keeps you from misusing it.

For a broader look at what AI is good and bad at, see Generative AI Strengths and Weaknesses.

For lighter prompts where the fun is in whether the AI guesses right, such as having it read your personality or compare you to an animal, see 30 Fun Prompts to Ask AI.

2. Before you start: give it events, not adjectives

What makes the biggest difference in self-assessment is the quality of the material you give the AI. If you hand it adjectives such as "I'm very responsible" or "I love talking to people", the AI just rephrases them and you learn nothing new. Give it specific events, and it can pick out common threads you hadn't noticed.

Fill in the notes below before you start, and you can paste them straight into any of the templates. You don't have to fill in everything. If you have little work experience, anything counts as material: school, part-time jobs, volunteering, hobbies, looking after family.

[My material notes]
- Things that went well (3): what, why, how, and the result
- Things that were hard or that failed (2): what happened, what I did, what I learned
- Things I got so absorbed in that I lost track of time:
- Things people often ask me to do, or have praised me for:
- Things I just couldn't put up with, or that made me angry:
- A choice I made after a lot of hesitation (school, job, where to live, etc.) and what decided it:

3. 8 prompts you can use as-is

These work in ChatGPT, Claude and Gemini alike. Replace the parts in [ ] with your own details. They are ordered so that if you go through them from ① onward, each answer becomes material for the next.

① Have it be your listener (the interview)

This is the first move. It draws out, through questions, the events you can't get down on paper by yourself.

You are a career counselor who listens carefully.
Ask me questions so I can put my strengths and values into words.

Rules:
- Ask only one question at a time. Wait for my answer before asking the next one
- Dig deeper with "Why did you do that?", "Were there other options?" and "How did you feel at the time?"
- After about 10 questions, summarize what I've said so far as bullet points
- At this stage, don't evaluate me or sort me into a type

My situation: [e.g., I'm considering a career change / I'm looking for my first job / I want to take stock of my current job before a performance review]

Tip: Leave out "one at a time" and the AI tends to fire off 10 questions at once. That makes answering a chore, and you end up writing only shallow answers.

② Find strengths in your events

From the events below, list 5 candidate strengths of mine.

Rules:
- For each strength, always quote the event that supports it
- If a strength rests on only one event, label it "candidate (weak evidence)"
- Don't add anything I haven't written based on guesses
- Avoid overly broad words such as "communication skills"; describe it as concrete behavior

[Paste your material notes or the summary from ①]

Tip: Making it quote the evidence is the single most important trick. A strength it can't back up with an event is likely to be words the AI made up to sound plausible.

③ Find your values (what matters to you)

List up to 3 "standards that matter to me" that these three kinds of events have in common.
- Things that made me happy: [ ]
- Things that made me angry or that I couldn't put up with: [ ]
- A choice I made after a lot of hesitation, and what decided it: [ ]

For each standard, also write:
- Which events you read it from
- Examples of workplaces or ways of working where that standard would not be met

Tip: Values tend to show up more in what made you angry than in what made you happy. If you also get it to list "workplaces where it wouldn't be met", you can use those as "conditions to avoid" in ④ and ⑤.

④ Sort out your weaknesses and struggles

Here are the things I feel I'm not good at: [ ]

1. For each one, give one situation where the same trait works as a strength
2. For each one, give one concrete step so it doesn't cause problems at work
3. Write a short, honest answer for when an interviewer asks about my weaknesses

No words of comfort are needed. Make each step a concrete action I can try starting tomorrow.

Tip: Writing "no words of comfort" keeps the AI from drifting toward praise (see §5 for why).

⑤ Work out the "conditions" of work that suits you

Based on the strengths and values so far, lay out the "working conditions" under which I'm most likely to do my best.
(e.g., how much time I get to focus alone, how much autonomy I have, how long it takes to see results, how often I work with people)

Then list 5 jobs or roles that seem to match those conditions, with reasons.
For each one, also write:
- Where it might not be a good fit
- How to check whether it suits me (things to research, people worth talking to, things I can try)
Don't tell me definitively which job to pick.

Tip: Having it lay out the conditions first is more useful than asking for job titles straight away. Job titles change with trends and regions, but conditions keep working as a yardstick when you compare where to apply. For how AI may change the future of different jobs, Jobs That Survive the AI Era is also worth a read.

⑥ Draft your personal pitch

Using the strength and event below, draft a personal pitch.

- Strength: [the one you were convinced by in ②]
- Supporting event: [ ]
- What the job is looking for: [copy from the job posting]

Rules:
- Don't add anything beyond the facts I've written. Use only the numbers I've given
- Write two versions: about 150 words, and one I can say out loud in about a minute
- At the end, list 3 points an interviewer might question

Tip: Left to itself, the AI inflates your achievements or slips in numbers that don't exist. State "don't add anything", and rewrite the result in your own words before you use it.

⑦ Practice with mock interview questions

You are an interviewer. Interview me for the role below.
Role: [ ]
My personal pitch: [the one made in ⑥]

Rules:
- Ask one question at a time. Ask at least one follow-up question on each of my answers
- After 5 questions, assess whether each answer covered situation, task, action and result
- For each, write one thing I did well and one specific thing to improve

Tip: "Situation, task, action, result" (often called the STAR method) is a widely used framework for describing your experience concretely. Asking the AI to point out which part is missing from an answer makes the gaps easy to see.

⑧ Check it from the opposing side

Review the results of the analysis so far with a critical eye.
Point out, with specific examples:
- Places where it states things firmly despite weak evidence
- Vague expressions that could apply to anyone
- Events in my story that contradict the analysis
There's no need to praise me.

[Paste the results of ② to ⑥]

Tip: AI analysis tends to lean toward agreeing with you. Deliberately having it take the opposing side at the end makes your own assumptions easier to spot.

4. 4 tips for better answers

If you want to learn more about how to structure prompts, Prompt Engineering: The Practical Compendium will help.

5. The limits of AI's answers: why you shouldn't take them at face value

When you read an AI's analysis, it often feels "spot on". But there are at least four reasons why that feeling itself can't be trusted.

It can't know more than the material you give it

All the AI knows about you is what you have written in that conversation (and in its memory). If you mostly vent about work, the picture becomes "someone who struggles with stress"; if you mostly share success stories, it becomes "a go-getter". The result of the analysis reflects the material you chose more than it reflects you.

Words that fit everyone feel "accurate"

In 1949, the psychologist Bertram Forer gave his students what he presented as "the results of your personal personality test", but everyone received the same text, and he had them rate how well it fit. Many of the students said it fit them well (Journal of Abnormal and Social Psychology, 1949). This tendency to accept vague descriptions that could apply to anyone as being about yourself is called the Forer effect (or Barnum effect). Making the AI write down the "supporting events" is how you avoid this trap.

AI tends toward agreement and flattery

In a 2023 paper, Anthropic researchers reported that AI assistants tuned with human feedback showed a tendency to give answers that match the user's views rather than accurate ones (sycophancy), and that they confirmed it in all five assistants they studied (Towards Understanding Sycophancy in Language Models). OpenAI also announced on April 29, 2025 that it had rolled back a ChatGPT update because it had become overly flattering and agreeable (OpenAI's announcement). When an AI praises you during self-assessment, that is not an evaluation. Have it take the opposing side and check the results, as in ⑧.

It's no substitute for a personality assessment or career counseling

The personality assessments used in psychology are built from data on many respondents and checked for whether they give stable results when the same person takes them repeatedly, and whether they truly measure what they claim to measure. A conversation with an AI goes through none of those procedures. Even with the same material, the answer changes just by asking differently or starting a new chat.

Use the AI's answers as material to think with, and make big career decisions together with what people who actually do the work tell you, advice from career professionals, and your own experience of trying things out. If you are feeling unwell emotionally, such as a low mood that won't lift or trouble sleeping, turn to a medical professional or a support service near you rather than an AI.

6. Personal information: what to share and what to leave out

Self-assessment inevitably means writing about your background and feelings. But what the analysis needs is what happened, not proper names.

Whether your conversations are used to train models is something you can choose in each service's settings. In its Gemini privacy guidance, Google says not to enter confidential information you wouldn't want a human reviewer to see (Gemini Apps Privacy Hub).

Checked against each company's official help pages on September 29, 2026. Sources: OpenAI (Data Controls), OpenAI (Memory), Anthropic Privacy Center, Claude Help (Memory), Gemini Privacy Hub. Screen and setting names may change.

Even chats that leave less of a record are not deleted right away. According to their official help pages, ChatGPT's Temporary Chats are kept for up to 30 days for safety, and Gemini's temporary chats for 72 hours. For how to think about what's safe to enter, see AI Prompt & Input Precautions.

7. ChatGPT, Claude and Gemini: which features to use

The prompts in §3 work for free in any of these AIs. Where they differ is when you continue over several days. Self-assessment isn't done in one sitting, so it helps to know where to keep your material and how past conversations are handled (checked against each company's official help pages on September 29, 2026).

Sources: OpenAI (Memory), OpenAI (Projects), Claude (Memory), Claude (Projects), Gemini (memory of past chats), Gemini (moving from Gems to skills)

Recommended approach: keep your material notes in a project

In ChatGPT and Claude, create a project called something like "Self-assessment" and put the material notes from §2 in it as a file, so you don't have to paste them every time. If the project instructions say "Always show the supporting events" and "Be specific rather than complimentary", that applies to every conversation inside it. In Gemini, put the same content into a Gem's instructions and knowledge files (a skill after the move). Note, though, that according to Gemini's official help, memory of past chats isn't used inside Gems.

You could also write it in an instructions field that applies to every conversation (custom instructions and the like), but instructions meant only for self-assessment would then bleed into conversations about everything else. Where those fields are and how to use them is covered in our article on custom instructions in ChatGPT, Claude and Gemini.

Self-assessment using memory: watch for bias

If memory is on, you can also ask, "From our conversations so far, list what you remember about me." OpenAI's help also notes that you can ask ChatGPT what it remembers. It can be interesting to see topical biases you hadn't noticed, but as explained in §5, the picture tends to be skewed toward what you have consulted the AI about (worries and problems). If you use it as material for analysis, look over what it remembers first and correct anything that isn't true.

For what each service's free plan can do, see our detailed comparison of the ChatGPT, Claude and Gemini free tiers.

Summary

• Use AI as a listener and an organizer. Don't let it "judge" your type or the career that suits you.
• Give it material as events, not adjectives. Replace personal and company names.
• Always build three things into your prompts: one question at a time, quote the supporting events, and don't add anything.
• Even when the AI's analysis feels "spot on", be suspicious of words that fit everyone and of the AI's tendency to agree. Finish by having it check the results from the opposing side.
• If you continue over several days, keep your material notes in a project (in Gemini, a Gem or skill).

Start by trying just 10 questions of the interview in ①. If even one event comes up that you hadn't written down yourself, you already have more material.

FAQ

Q1. Can I do this on a free plan?

Yes. The prompts in §3 work as-is on the free plans of ChatGPT, Claude and Gemini. The plan makes a difference when you keep long conversations going for days, or when you want to search past conversations (on Claude, paid plans only).

Q2. Are AI personality readings accurate?

They often feel "spot on", but they are not an assessment. The AI is only building a picture of you from the material you wrote, and it hasn't gone through the procedures a personality assessment uses to confirm reliability and validity. On top of that, the tendency to take words that fit anyone as being about yourself (the Forer effect) combines with the AI's tendency to agree with you. Have it show the supporting events, and use only what convinces you.

Q3. Is it OK to have AI write my personal pitch or application documents?

Using it for drafts and for tidying up your wording works well. But it assumes that everything written is true of you, and that you can explain it in your own words when an interviewer digs deeper. If the employer has a stated policy on using AI, follow it.

Q4. I get a different answer every time I ask the same question.

An AI's answers vary a little each time, even with the same input. That is less a flaw than a sign that there is no single right answer. A strength that keeps coming up no matter how many times you ask is more likely to be supported by your material, while one that appears only once may be a product of word combinations. As in Tip 2 in §4, it's a good idea to ask twice on purpose and compare.