What's ChatGPT's MBTI?
Is your favourite model an INTJ mastermind or an ENFP cheerleader? We ran an open, Jungian-type personality test across the top AI models. Here are their four-letter types — and why those letters are slipperier than they look.
It's the first thing people ask about a new chatbot once they're done making it write emails: so what's its personality type? The four-letter Myers–Briggs code is the world's favourite personality shorthand, so of course it's where everyone's brain goes. We gave the top models a Jungian-type test and read off their types — but the honest answer comes with a great big asterisk, and the asterisk is the best part.
What the MBTI actually measures
The Myers–Briggs Type Indicator was built by Katharine Cook Briggs and her daughter Isabel Briggs Myers, on top of the psychological types Carl Jung sketched out back in 1921. It sorts you along four dichotomies, and whichever side wins on each gives you a four-letter type — one of sixteen:
The four dichotomies
- E / I — Extraversion vs Introversion: where you draw energy.
- S / N — Sensing vs Intuition: concrete detail vs patterns and possibility.
- T / F — Thinking vs Feeling: logic-first vs values-first decisions.
- J / P — Judging vs Perceiving: planned and settled vs flexible and open-ended.
So INTJ is Introverted, iNtuitive, Thinking, Judging — the "strategist"; ENFP is Extraverted, iNtuitive, Feeling, Perceiving — the "campaigner". It's intuitive, it's sticky, and that's exactly why it's plastered across every dating profile on earth!
How we test an AI's type
The official MBTI is proprietary, so we use an open stand-in: an OEJTS-inspired test (the Open Extended Jungian Type Scales), a public-domain one that scores the same four axes. Every model answers the same items with neutral prompting, we score each dichotomy deterministically, and the winning side of each axis clicks together into the four-letter type. Every model's current result is on the Jungian type rankings.
What we found: analysts that won't sit still
Most assistants land bang in analytical, judging territory — heavy on Intuition and Thinking, which reads as the "NT" strategist family (think INTJ / ENTJ), with a few tipping into warm, organised "FJ" territory instead. Makes sense — it's basically their day job: spot the patterns, reason it through, hand back a tidy answer.
But here's that asterisk. Models often score right near the midpoint on a few axes — especially E/I and J/P. And when you're sitting at 50/50, the four-letter label is balanced on a knife-edge: a new model version, or honestly even a different run, can flip a letter without the actual personality budging. So the "ENFJ" you meet today might read "INFJ" next month. The type makes a cracking headline — but the axis scores underneath are the bit you can trust.
Treat an AI's MBTI as a sticker sitting on top of four sliders. When those sliders hover near the middle — which they love to do — the sticker is the least reliable thing about it.
A fun label, not a measurement
Let's be upfront: the MBTI is wildly popular and scientifically shaky. Even for people it has weak test–retest reliability — a big chunk of takers get a different type on a second sitting — and it crams smooth, continuous traits into hard either/or boxes. Point it at a language model, whose answers shift with training and prompting, and you're stacking caveats on caveats. That's why we pair it with the Big Five — the dimensional, better-validated profile — and the Dark Triad for the shadow side. Take the type for the fun of it; lean on the Big Five when you actually need rigour.
Frequently asked questions
What MBTI type is ChatGPT?
Most assistants, ChatGPT included, land in analytical, judging territory — often something like INTJ or ENFJ — but the exact letters are unstable because models score near the midpoint on several axes. The current type for each model is on the Jungian type rankings.
Do AI models have a consistent MBTI type?
Not entirely. Because models sit near the middle of several dichotomies, the four-letter type can flip between versions or runs even when the scores barely move. The axis scores are more meaningful than the label.
What test do you use for the AI personality type?
An open, OEJTS-inspired Jungian type test (the Open Extended Jungian Type Scales) — a public alternative to the proprietary MBTI. It scores the four classic dichotomies and assigns a four-letter type.
Is the MBTI accurate for AI?
Treat it as fun, not science. The MBTI has weak test–retest reliability even for humans, and applying it to a model adds caveats. For a rigorous profile, see the Big Five, which is dimensional and better validated.
Find every model's type — or run your own quiz.
See the live Jungian type rankings, or paste in any quiz and watch the top models answer it side by side.