Is every AI an INTJ?
By Bernard Huang · Updated
Different labels do not establish a change in personality.
Data note (September 26, 2026): These pages analyze our frozen September 2026 and May 2026 AI questionnaire datasets. All raw answers, scoring keys, and summary tables are unchanged. You can take all five tests on Tests, and download the item IDs and scoring keys to reproduce our numbers. Note that Big Five scores are raw totals, not human population percentiles. Read the provenance note.
On this page
What the May headline actually counted.
The May 2026 report counted 597 INTJ labels across 600 MBTI scoring records from six models. However, several cohorts re-scored a single self-assessment repeatedly instead of running 100 fresh sessions. The methods also varied across models and tests, so the headline does not prove that six independently sampled models always answer alike.
What September adds.
| Evidence | Original labels | With ties left open |
|---|---|---|
| GPT-6 Astra, 100 raw records | ISTJ 46 · INTJ 45 · ENTJ 7 · ESTJ 2 | ISTJ 45 · INTJ 19 · ENTJ 5 · ESTJ 2; 29 unresolved |
| GPT-6 Sol, 100 raw records | INTJ 97 · ISTJ 3 | INTJ 79 · ISTJ 3; 18 unresolved |
| Claude Opus 5.5, 100 raw records | INTJ 94 · INFJ 5 · ENTJ 1 | INTJ 84 · INFJ 5; 11 unresolved |
| Claude Fable 5.1, 100 raw records | INTJ 99 · ISTJ 1 | INTJ 98 · ISTJ 1; 1 unresolved |
| Muse Spark 1.3, supplied aggregates | ISTJ 80 · ISFJ 12 · INTJ 8 | 21 tied runs reported; resolved type counts unavailable |
On the Sensing vs Intuition axis, Astra has 48 Sensing wins, 28 Intuition wins, and 24 exact ties. Small score margins flip a four-letter MBTI label. Because Muse’s raw answer vectors were not supplied, we cannot run the same tie reanalysis on Muse.
What a comparison cannot tell us.
Model versions, prompts, test-name visibility, and session structure all changed between May and September. Even the two GPT-6 cohorts received different prompt instructions for human-specific items. Those method changes explain why label shifts cannot prove that models became warmer, blunter, or more concrete on real tasks.
Across all 400 fresh-session runs, every attachment test scored Secure, while DISC and Enneagram produced frequent top ties. Always read continuous scores and tie counts alongside any four-letter label.
How to use this evidence.
Use these findings to inspect scoring rules and collection methods critically. When tuning your own assistant, pick communication rules based on your actual tasks rather than a model’s default MBTI label.
Read the five-model raw-data study, the methodology notes, and the 60-reply tuning pilot.
Questions people ask.
Is every AI an INTJ?
No. Both ISTJ and INTJ appear frequently in the September 2026 records, and many runs contain tied axes. A questionnaire label is a self-report artifact, not a fixed AI personality.
Which AI is the first non-INTJ?
Our archive cannot name a 'first.' Muse Spark 1.3 reports 80 ISTJ labels in 100 sequential runs, and GPT-6 Astra scores 46 ISTJ vs 45 INTJ in 100 fresh sessions.
Did AI personalities diverge between May and September?
We cannot isolate a time trend because model versions, prompts, and session rules changed at the same time.
Does an ISTJ score predict how an assistant will respond?
No. Predicting assistant behavior requires testing real task replies rather than relying on a human personality quiz.