Real test distributions
How common is your personality type, really? These are honest, first-party result distributions computed from 113,370 completed JobCannon assessments across 18 tests — not textbook figures or community polls.
- 18
- Assessments
- 113,370
- Completed tests
- 117
- Result types
- 12
- Languages
Published type frequencies are usually inherited: a normative sample collected decades ago, quoted by one source, then quoted from that source until the original sample is three citations away and nobody re-counts it. These are counted. Every share on every page below comes from the live result table, recomputed on a fixed schedule and stamped with the date it was pulled — 2026-08-16 for the figures you are reading now.
The unusual part is the 12 languages. 18 of the 18 tests here were answered in more than one, with the same items and the same scoring code on every side, which makes the columns comparable in a way that separately-run national studies never are. Each test page states how far its most common result actually travels between its highest and lowest language.
Jungian Archetype
n = 32,171- Most common
- Creator (15.8%)
- Rarest
- Ruler (3.3%)
- Result types
- 12 · 6 languages
MBTI Personality
n = 19,072- Most common
- ISTJ — The Logistician (23.9%)
- Rarest
- ENTJ — The Commander (1.6%)
- Result types
- 16 · 10 languages
Spirit Animal
n = 19,052- Most common
- Wolf (25.9%)
- Rarest
- Lion (3.7%)
- Result types
- 8 · 9 languages
DISC Profile
n = 5,830- Most common
- Steadiness (27.5%)
- Rarest
- Influence (19.8%)
- Result types
- 4 · 5 languages
Enneagram
n = 4,886- Most common
- Type 5 — The Investigator (24.1%)
- Rarest
- Type 9 — The Peacemaker (3.0%)
- Result types
- 9 · 7 languages
8 Thinking Styles
n = 4,470- Most common
- Interpersonal Intelligence (19.0%)
- Rarest
- Naturalist Intelligence (4.9%)
- Result types
- 8 · 4 languages
Big Five Personality
n = 4,297- Most common
- Openness (36.7%)
- Rarest
- Extraversion (5.5%)
- Result types
- 5 · 4 languages
Learning Style Test
n = 3,491- Most common
- Visual (45.0%)
- Rarest
- Kinesthetic (12.0%)
- Result types
- 4 · 3 languages
Temperament
n = 3,291- Most common
- Sanguine (28.7%)
- Rarest
- Choleric (19.2%)
- Result types
- 4 · 3 languages
EQ
n = 3,248- Most common
- Self-Awareness (42.3%)
- Rarest
- Social Skills (12.1%)
- Result types
- 4 · 4 languages
Love Languages
n = 2,873- Most common
- Acts of Service (31.2%)
- Rarest
- Quality Time (9.1%)
- Result types
- 5 · 7 languages
Moral Alignment
n = 2,448- Most common
- Neutral Evil (18.9%)
- Rarest
- Chaotic Good (5.2%)
- Result types
- 9 · 2 languages
Communication Style Test
n = 1,679- Most common
- Assertive (64.1%)
- Rarest
- Aggressive (9.5%)
- Result types
- 4 · 2 languages
Type A / Type B Personality
n = 1,637- Most common
- Type B (40.1%)
- Rarest
- Type A (22.5%)
- Result types
- 3 · 3 languages
Conflict Styles
n = 1,321- Most common
- Collaborating (50.4%)
- Rarest
- Avoiding (4.6%)
- Result types
- 5 · 3 languages
Dark Triad
n = 1,320- Most common
- Machiavellianism (52.8%)
- Rarest
- Psychopathy (7.2%)
- Result types
- 3 · 3 languages
Values Assessment
n = 1,274- Most common
- Security (22.4%)
- Rarest
- Recognition (0.9%)
- Result types
- 10
Attachment Styles
n = 1,010- Most common
- Anxious (37.5%)
- Rarest
- Fearful-Avoidant (10.5%)
- Result types
- 4 · 3 languages
Download the raw dataset
Item-level anonymized responses — every answer, dimension score, and result type, with no names, emails, or identifiers of any kind. Released under CC BY 4.0 for researchers, educators, and benchmarking, and mirrored across 7 services so a dead host never takes the citation with it — 6 of them already carry the current cut.
The download is a frozen v3 snapshot cut on 2026-08-15: 54,431 responses across 9 instruments, of which 16,112 are non-English across 24 languages — the same items under the same scoring, which is the part open psychometric corpora almost never have. That is deliberately not the same corpus as the aggregates above, which cover 18 tests and keep moving — a citation needs something that will still say the same thing next year, and a percentage on a live page cannot promise that.
| MBTI Personality | 17,761 |
| Career Match | 8,445 |
| RIASEC Career Interest | 5,891 |
| DISC Profile | 5,444 |
| Enneagram | 4,412 |
| 8 Thinking Styles | 4,161 |
| Big Five Personality | 3,997 |
| EQ | 3,100 |
| Dark Triad | 1,220 |
| Total | 54,431 |
Harvard Dataverse (DOI) still serves v2 — 44,342 responses cut on 2026-08-06. The handle is permanent and will resolve to v3 once it is deposited; until then cite the version you actually downloaded rather than the total above.
What this data is not
Everyone counted here chose to take a personality test on the internet. That is a real population and an interesting one, but it is not a random sample of any country, workforce, or age group. Read every share as how results are distributed among people who take these tests, which is the honest claim, rather than as a national frequency — the second reading is where inherited type statistics usually went wrong in the first place.
There is no demographic split here because we do not collect the demographics. No age, no gender, no country, no occupation — so no cut by any of them, and we would rather publish nothing on that than infer it from a name or an IP address. The language columns are the interface language a person answered in, which correlates with country but is not the same thing.
None of this is a clinical or diagnostic instrument, and nothing here is a hiring recommendation. Reliability figures for the tests that have them are on the reliability page — a well-behaved distribution says nothing on its own about whether an instrument measures what it claims to.
How to cite this page
The aggregate distributions are published under CC BY 4.0. Reuse them anywhere with attribution. Cite this page for the aggregates and the DOI for the raw file — they cover different corpora, and a reader who follows the wrong one will not find the number you quoted.
JobCannon (2026). Real personality test distributions across 18 assessments (n = 113,370) [Data set]. JobCannon Research. Retrieved 2026-08-16, from https://jobcannon.io/research/distributions
113,370 completed assessments across 18 tests, pulled 2026-08-16. The largest single test is Jungian Archetype at 32,171 takers. Everything published here is aggregate — result types and dimension scores only, never an individual answer, name, or email. A test appears once it clears 100 completed results, and a language column appears once that language clears 100 on its own, so a small cell is left out rather than published as a percentage of almost nothing.
Most published type frequencies trace back to a small number of mid-century normative samples, republished for decades without recollection. These are counted directly from 113,370 people who took the test on one platform, with one scoring implementation, and are recomputed as volume grows. Where the two disagree, that is usually a real difference between two populations — an online self-selected audience is not a stratified national sample — rather than a correction to the other source.
Yes, and that is the part of this corpus nobody else publishes. 18 of the 18 tests carry a per-language breakdown covering 12 languages — English, Japanese, Spanish, Arabic, Indonesian and others — with identical items and identical scoring, so the columns are genuinely comparable rather than separate studies stitched together. Each test page shows how far its most common result moves from the highest-share language to the lowest.
Yes. 54,431 item-level anonymized responses across 9 instruments are published under CC BY 4.0, of which 16,112 are non-English across 24 languages — identical items, identical scoring, so the languages are comparable rather than separate studies. That download is a frozen v3 snapshot from 2026-08-15, live on Hugging Face, Kaggle, GitHub, Zenodo (DOI), Figshare (DOI), OSF. Harvard Dataverse (DOI) still serves v2 (44,342 responses) until it is re-deposited, so check the version before you cite a row count. The aggregate percentages on this page cover a larger, continuously refreshed corpus again, so none of them will match row for row.
Yes, under CC BY 4.0 with attribution. Cite the page for the aggregate distributions and the DOI for the raw file — they are different corpora. Harvard Dataverse (DOI) still resolves to v2, so state the version alongside the handle. Suggested form: JobCannon (2026). Real personality test distributions across 18 assessments (n = 113,370) [Data set]. JobCannon Research. Retrieved 2026-08-16, from https://jobcannon.io/research/distributions
The aggregates do. They are recomputed from the live result table and stamped with a pull date — this page reads 2026-08-16. Shares shift by fractions of a point between pulls as more people take each test, so quote the pull date alongside the figure. The DOI-minted snapshot is the opposite: it never changes, which is exactly why it is the thing to cite in a paper.
Every figure is aggregate-only — result type and dimension scores, no names, emails, or individual answers. Numbers refresh as more people take each test. Pulled 2026-08-16. For our research methodology, see the methodology page.