There are thousands of personality tests online, and from the outside they all look alike: a clean progress bar, some statements to agree or disagree with, and a result written with great confidence. A few are built on decades of research. Many are built on nothing at all. Very few of them make it easy to tell which is which.
We run a site full of these tests, so you should read this with that in mind — we have an interest in you trusting tests, and an interest in you trusting ours in particular. We have tried to write the checklist we would want a friend to use on us. Some of the checks our tests pass comfortably. One or two are places where we have made trade-offs you are entitled to know about.
None of this requires a statistics background. Most of the checks take seconds, and the first three will rule out the majority of what you find.
The Big Five, also called the Five-Factor Model or OCEAN, describes personality along five independent dimensions: Openness, Conscientiousness, Extraversion, Agreeableness, and Neuroticism. This test administers the IPIP-50, a fifty-item public-domain scale developed by Lewis R. Goldberg, and reports a score on each dimension rather than assigning a single personality type.
1. Can you find out where the questions came from?
This is the single most telling check. A serious test will tell you which instrument it uses — a named questionnaire, who developed it, and ideally a link to the source. A test that only says "developed by our team of experts" or "based on psychology" is telling you, in effect, that you cannot check.
There is a reason this matters beyond transparency. Writing good personality items is hard. They need to be tested on large samples, checked for whether they actually hang together, revised, and tested again. Public item pools exist precisely so that people do not have to start from scratch. Our Big Five test, for example, uses the IPIP-50, a fifty-item public-domain scale from the International Personality Item Pool based on Lewis Goldberg's 1992 factor markers, and we reproduce the items verbatim and link to them. Our career interests test uses the U.S. Department of Labor's O*NET Interest Profiler. You can open the source and compare.
If you cannot find the source after a minute of looking, treat the result as entertainment. It might be fun. It is not information.
2. Does it say what it is not?
Good tests are surprisingly keen to tell you their limits. A career interest test should tell you it measures interest, not ability. A four-letter type test that is not the official MBTI should say so, because the MBTI is a trademarked commercial instrument and an independent test that implies otherwise is misleading you about something basic. A personality test should say it is not a clinical assessment and not a hiring tool.
A test that claims to reveal your "true self", predict your perfect career, or diagnose anything is overselling. The more a test promises, the less you should believe it.
3. Does it report scores, or only boxes?
People are not distributed in neat groups. Most traits are spread continuously, with most people somewhere in the middle. A test that reports where you sit on each dimension is describing that honestly. A test that only hands you a type, with no scores behind it, has thrown information away — and you have no way of knowing whether you are a strong example of your type or one answer away from a different one.
This is the main reason the Five-Factor Model is the dominant framework in academic personality research: it reports five continuous scores rather than forcing each dimension into two boxes, and its scales show stronger reliability and predictive validity than type sorting. Type tests are not useless — they are popular because a four-letter label is easy to talk about — but a good one shows you the scores underneath and tells you when a letter was close.
4. Does it admit how precise it is?
Every questionnaire score contains measurement error. A test that knows this will show it: a likely range around your score, a note when two of your scores are too close to rank, or a flag on any letter that could flip on a retake. A test that presents every number as exact, to the percentage point, is either unaware of its own error or hoping you are.
This is the check we care most about, and the one we have put the most work into on our own result pages. When the data exist, we compute the error from the instrument's published reliability and show it. When they do not, we say so and use a stated rule instead. It makes results look less impressive. We think it makes them more trustworthy.
5. Who are you being compared with?
"You are more extraverted than 80% of people" only means something if you know who the people are. Percentiles come from a reference sample, and a good test tells you what that sample was. A test that gives percentiles with no stated sample has probably invented them, or compared you only against other visitors to the same site without telling you.
Even honest samples have limits worth knowing. The reference sample behind our type test is over 230,000 people who answered the same items online — large, but self-selected and mostly North American. That is a reasonable comparison group for an online test and a poor one for, say, claims about a whole country. The best tests tell you this themselves. If a test is coy about its norms, be sceptical of its percentiles.
6. Can you agree with everything and get a sensible result?
A quick experiment you can run on any test: think about what would happen if someone simply agreed with every statement. In a well-built questionnaire, nothing very interesting — because about half the statements for each trait are worded in the opposite direction, so agreeing with everything cancels out. "I am the life of the party" and "I don't talk a lot" sit on the same scale, scored in opposite directions.
If every statement in a test is phrased the same way round — all of them describing the trait as present — then a person who tends to agree will score high on everything. That is a design flaw, and it is surprisingly common. Scan the items. If you cannot find any reversed ones, the test is measuring agreeableness with the test as much as anything else.
Some tests go further and check the quality of your answers directly: answering faster than the statements can be read, giving the same response to almost every item, or agreeing with a statement and its opposite. Ours do this and tell you if a result looks affected. It never blocks or changes your result — that would be presumptuous — but it is the kind of thing a careful test notices.
7. Is the result flattering in a way that would fit anyone?
Read your result and ask honestly: would this paragraph fit most of my friends too? "You value deep connections but need time alone to recharge." "You are driven, but sometimes doubt yourself." Psychologists call the tendency to accept vague, flattering descriptions as personal the Barnum effect, and a lot of online tests run on it.
A useful test says things that could be wrong about you. It tells you that you scored low on something, or that two of your traits pull in different directions, or that a result was too close to call. If every type, every score, and every profile reads as a compliment, the test is optimised for sharing, not for accuracy.
8. What does it ask for, and when?
Look at what the test wants from you before it shows a result. A name and email address before you can see anything usually means the result is the bait and your contact details are the product. A demand for your phone number, date of birth, or employer has nothing to do with measuring personality.
Money is a fairer question than it first appears. We sell a paid report, so we will not pretend that charging for anything is a red flag. But there is a meaningful difference between a test that shows you a real free result and offers more detail for a fee, and one that hides every result behind a paywall after you have invested twenty minutes. In our case the free result gives you your type or code and some of the scores behind it; the remaining scores and the longer write-up are in the paid report. That means a free result shows fewer scores than check 3 would ideally like, and we would rather say so than pretend otherwise. You can decide whether the trade is fair — but you will have seen your result before you decide.
9. Does it overreach on what the result is for?
Personality tests are good for reflection, conversation, and self-understanding. They are much weaker as tools for making decisions about other people. A free online test that suggests you use it to screen job candidates, pick a partner, or judge a child's potential is overreaching. Even research-grade instruments are used carefully in those settings, by trained people, alongside other evidence.
The same applies to claims about what your result means for your life. "People with your type should become engineers" is a jump no questionnaire can support. Interest and personality can point you toward options worth exploring. They cannot tell you which one to choose.
10. Would someone who knows you recognise the result?
The last check is the one no website can do for you. Show your result to a person who knows you well and ask if it sounds like you. If they laugh and say "that is exactly you", that is meaningful evidence. If they look puzzled, it is worth taking seriously — either the test missed, or you answered as the person you would like to be rather than the person you are.
Our Big Five test makes this formal: after your result you can send a link to someone who knows you, they rate you on the same statements, and the page shows where your view and theirs differ by more than measurement error. Self and others tend to disagree in informative ways. You do not need a website for the informal version, though. A five-minute conversation will do.
What a good test still cannot do
Passing all ten checks does not make a test an oracle. A well-built, well-normed, honest questionnaire still only knows what you told it, on one day, in answer to a few dozen statements. It describes tendencies, not destiny, and it will be wrong about some things for some people.
That is not a reason to dismiss personality tests. It is a reason to choose them carefully and hold the results lightly. Used that way, a good one can give you language for things you half-knew about yourself, and a few things you did not. If you want to start with something that passes the checklist, the Big Five is the most researched framework there is, and our guide to the Big Five traits explains what each score means.