Are online Big Five tests reliable?
This page answers a simple question: can you trust a personality questionnaire taken on a screen? It draws on three studies whose abstracts only were read, and it says what each one lets us claim, and what it does not.
1What does “reliable” mean for a personality test?
Three different things hide behind the word. Does the questionnaire give roughly the same result each time? Does it measure what it says it measures? And does it compare you with other people, or only with a scale? A test can be good on one of these and weak on the other two.
| The question | The technical term | What you can check yourself |
|---|---|---|
| Is the result roughly the same each time? | Reliability (stability of the result) | Whether the site publishes a repeat-testing study. If not, retake the test later and compare |
| Does the questionnaire really measure the five-factor model? | Validity | Where the items come from, and whether a published study compared them with anything else |
| Does my score compare with other people? | Norming | Whether the site says which group of people the scores were calibrated on. If not, the score compares with no one |
Most results pages answer the third question without asking it, by showing a percentage that looks like a rank. A percentage of points is not a rank, and the page explaining the method should say so.
2Does answering on a screen change the result?
The available data say very little. In 2004, Gosling and colleagues compared a sample of 361,703 online responses with 510 published samples collected by traditional methods. Their conclusions, according to the abstract: online respondents are fairly diverse in gender, socioeconomic status, region and age; findings hold across presentation formats; careless responses and repeat responders do not distort the conclusions; and the results are consistent with those from traditional methods.
That study is about collecting data for research, not about the quality of any particular website. It covers people who choose to take part. It shows that the screen is not the problem. It does not say that a questionnaire found online is a good one.
3Do you get the same result twice?
Roughly, but not exactly, and the gap grows with time. In 2000, Roberts and DelVecchio pooled the results of 152 studies that follow the same people over time. For an interval of 6.7 years, the correlation between two testings is .54 in college students, .64 at age 30 and .74 between ages 50 and 70.
| Age | Test-retest correlation |
|---|---|
| College years | .54 |
| Age 30 | .64 |
| Between ages 50 and 70 | .74 |
Two things follow. First, a correlation of .54 to .74 is clear but not total: people keep roughly their place relative to one another, and their scores still move. Second, the abstract notes that the longer the interval, the lower the correlation. These figures cover six and a half years, not a week, and they describe groups: they do not tell you by how much your own score will change.
In practice, two sittings a few weeks apart will not give exactly the same number, and that is not a sign of a faulty test. Your mood that day, the setting you have in mind while answering (work, close ones) and tiredness all play a part. A gap of a few points between sittings reads as noise, not as change.
4How do you recognise a serious questionnaire?
By what it says about itself. Four points can be checked in two minutes on the method page, without knowing any statistics. The table compares them with what our own questionnaire does.
| What to look at | What a serious questionnaire does | Our questionnaire |
|---|---|---|
| Where the items come from | It is written down: a published inventory, a free item bank, or items written for the purpose | Items written for Brain Quiz from the literature. No published inventory is reproduced |
| Balance of wording | Items worded both ways, so that someone who agrees with everything is not rewarded | Six items per factor: three straight, three reversed |
| Norming | It is stated: which population the scores were calibrated on, or an admission that they are not | No norming. The percentage shown is a share of points, not a rank |
| Level of detail | It is announced: five scores only, or facets under each factor | Five scores, no facets |
The second point comes from a known difficulty: some people tend to agree whatever the statement says. Soto and John, in the 2017 article presenting the BFI-2, note that their inventory controls for this tendency. They also organise the model into five domains and fifteen facets, and conclude, according to the abstract, that the inventory is reliable and valid. The BFI-2 is protected by copyright: we use none of its items.
Thirty items do not give the fineness of a longer inventory. The price is precision: the shorter a questionnaire, the less well it tells two close profiles apart. Our questionnaire is a proprietary self-observation questionnaire written from the literature. It is not normed, and it can neither confirm nor rule out a condition.
5How do you read your score without over-reading it?
- Read it as a snapshot. Your answers reflect what you think of yourself today, not a fixed trait.
- Compare yourself with yourself, not with others. Without norming, a score of 70 out of 100 does not mean “higher than 70% of people”.
- Do not look for the good side. None of the five factors is better than another; the page “The Big Five model explained” details what each end describes.
- Retake the test in a few months, at a calm moment. A gap of a few points is noise; a clear gap on one factor deserves a look.
- Ask someone close how they would describe you. An outside view completes what people say about themselves.
What this guide does not say
This page draws on three studies whose abstracts only were read; it takes from them only what those abstracts say. It does not compare existing Big Five inventories one by one, does not put a figure on the reliability of any particular website and does not cover the use of these tests in hiring. The correlations cited are for groups, years apart, not for your own result from one week to the next. It does not replace professional advice.
Frequently asked questions
- Is a free Big Five test worth less than a paid one?
- Price says nothing about quality. None of the studies cited here compares free tests with paid ones. What counts can be read on the method page: where the items come from, the balance between straight and reversed items, norming and the stated limits. A test that publishes none of this cannot be assessed, free or not.
- How many questions does a Big Five test need?
- There is no magic number. A short questionnaire is quick and separates close profiles less well; a long one is finer but tiring. Ours has 30 items, six per factor, for about eight minutes. It gives a first picture in five scores, not a detailed portrait by facets.
- Can you take the test more than once?
- Yes, and it is useful for seeing what moves. Stability studies cover groups over several years: a correlation of .54 to .74 over 6.7 years depending on age. Between two sittings, small gaps are normal. Compare your own results rather than chasing an exact figure.
- Is the result the same on a phone and on a computer?
- The abstract of Gosling and colleagues' 2004 study says findings hold across presentation formats. That study is not about our site: the main point is that the screen is not the main source of gaps, unlike mood and the setting you have in mind while answering.
- Can a Big Five test tell me whether I have a condition?
- No. The Big Five describes behavioural tendencies found in everyone, to varying degrees. No personality questionnaire can confirm or rule out a condition. Only an assessment carried out by a health professional can address that question.
Sources
- Gosling, S. D., Vazire, S., Srivastava, S., & John, O. P. (2004). Should we trust web-based studies? A comparative analysis of six preconceptions about internet questionnaires.. American Psychologist, 59(2), 93-104. DOI 10.1037/0003-066X.59.2.93 (opens in a new tab)Un échantillon de 361 703 réponses en ligne comparé à 510 échantillons publiés par des méthodes classiques. Seul le résumé a pu être lu. Consulte le 2026-11-06.
- Roberts, B. W., & DelVecchio, W. F. (2000). The rank-order consistency of personality traits from childhood to old age: a quantitative review of longitudinal studies.. Psychological Bulletin, 126(1), 3-25. DOI 10.1037/0033-2909.126.1.3 (opens in a new tab)Méta-analyse de 152 études suivies dans le temps. Corrélations test-retest de .54 à l'université, .64 à 30 ans, .74 entre 50 et 70 ans, pour un intervalle de 6,7 ans. Seul le résumé a pu être lu. Consulte le 2026-11-06.
- Soto, C. J., & John, O. P. (2017). The next Big Five Inventory (BFI-2): Developing and assessing a hierarchical model with 15 facets to enhance bandwidth, fidelity, and predictive power.. Journal of Personality and Social Psychology, 113(1), 117-143. DOI 10.1037/pspp0000096 (opens in a new tab)Un inventaire en cinq domaines et quinze facettes, qui maîtrise la tendance à répondre « d'accord » quel que soit l'énoncé. Seul le résumé a pu être lu. Consulte le 2026-11-06.
Related questionnaires
These questionnaires cover the subject of this guide. Each one states where its statements come from and how it is scored.
Big Five personality test
Free online Big Five personality test, no sign-up: 30 items, the five factors computed separately, and the scoring method published.