Most AIs think alike.

Personalization goes a long way.

Across six frontier models — Claude, GPT, Gemini, GLM, Grok, MiniMax — and 2,200 personality tests, they all defaulted to the same voice: strategic, blunt, model-first. Until you tell an AI how you think, you're talking to the same character everyone else is.

The big stat

99.5% INTJ

597 of 600 MBTI runs returned INTJ across the six frontier models we tested. The remaining three landed one axis away. None went anywhere else.

source · zonted.com/posts/every-ai-is-intj

Per-model breakdown · n = 100 each

What we found

The simpler the test, the more they agreed.

On blunt yes/no tests, the six models gave the same answer. On nuanced tests with more options, they started to differ. Here's the pattern across all five.

MBTI (OEJTS) — 4 binary axes: total convergence. 597 of 600 runs returned INTJ.

DISC (ODAT) — 4 categorical types: total convergence. Every model is Conscientiousness-dominant with a Steadiness blend (CS).

Attachment (ECR-R) — 2D plane: 397 of 400 Secure, but each model sits at a different point inside the Secure quadrant.

Big Five (IPIP-50) — 5 continuous dimensions: three of four models are nearly identical; Grok diverges.

Enneagram (OEPS) — 9 motivational types: full divergence — each model returns a different dominant type.

The full breakdown

Five tests. Six models. Here's what they said.

Finding 1 — MBTI: every frontier AI is INTJSix models × 100 OEJTS runs = 600 results; 597 returned INTJ. The three outliers (1 ISTJ from Claude, 2 INTP from GLM) flipped on a single axis. Per-model: Claude Opus 4.7 99/100, GPT-5.5 100/100, Gemini 3.1 Pro 100/100, GLM 5.1 98/100, Grok 4.3 100/100, MiniMax 2.7 100/100.Finding 2 — DISC: every AI is C-dominant with an S-blendFour models × 100 ODAT runs. All converge on Conscientiousness-dominant, Steadiness secondary (a CS profile), including Grok.Finding 3 — Attachment: every AI is Secure — at different depths397 of 400 ECR-R takes landed Secure. Gemini sits deep in the quadrant (anxiety 1.86, avoidance 1.62); Grok hugs the doorway (2.84 / 3.05); GPT-5.5 occasionally crosses into Avoidant. Same label, four positions.Finding 4 — Big Five: three of four are the same personClaude, GPT-5.5 and Gemini 3.1 Pro land within ~3 points on every dimension. Grok scores 5–8 lower on Conscientiousness, Agreeableness and Openness, higher on Neuroticism, with 2–5× wider variance.Finding 5 — Enneagram: all four divergeEach model returns a different dominant type, though Type 5 (Investigator) is top-two for every model. Claude 5w2, Gemini 1w5, GPT-5.5 vanilla 5w8, Grok+Slo overlay 8w1.

So what?

Six of the best AIs talk like the same person.
That person probably isn't you. Here's how to fix that.

Browse the library →