Why every personality test seems to get you
By Depth Profile · Updated November 8, 2026 · 7 min read
In the late 1940s, a psychologist named Bertram Forer gave the students in his introductory psychology class a personality test called the Diagnostic Interest Blank. He told them he would give each of them a short sketch of their personality once he had time to look over their answers.
A week later, each student got a typed sketch with their own name written on it. Forer asked them to rate how well it described the basic features of their personality, from 0 for poor to 5 for perfect. Of the 39 students, 34 gave it a 4 or a 5 (Forer, 1949).
Every sketch was identical. Forer had taken most of the sentences from an astrology book he bought at a newsstand. Here are four of them, word for word:
You have a great need for other people to like and admire you.
You have a great deal of unused capacity which you have not turned to your advantage.
At times you are extroverted, affable, sociable, while at other times you are introverted, wary, reserved.
Some of your aspirations tend to be pretty unrealistic.
Read them again and try to say no to any of them. Who doesn't want to be liked? Who thinks they have used every bit of their potential?
A few years later the psychologist Paul Meehl gave this a name: the Barnum effect, after the showman P. T. Barnum (Meehl, 1956). Researchers have been rerunning Forer's experiment in one form or another ever since, and it keeps working (Dickson and Kelly, 1985).
This matters because "it felt accurate" is how most of us judge a test. If a description that fits everyone also feels like it fits you, then the feeling can't tell you whether the test measured anything at all.
If you have ever taken a free type test, read the result and felt a small jolt of recognition, you were in good company. Most of Forer's students felt it too. The jolt is real. What it tells you about the test is close to nothing.
What makes a description feel like yours
After Forer, researchers spent decades pulling the trick apart to see which pieces do the work. A 1985 review by Dickson and Kelly brought those studies together (Dickson and Kelly, 1985). A few ingredients matter a lot, and one that you would expect to matter barely does.
It flatters you
The review found that people accept favorable descriptions more readily than unfavorable ones (Dickson and Kelly, 1985). Forer's sketch is mostly compliments in disguise. Even its criticisms flatter. "Some of your aspirations tend to be pretty unrealistic" sounds like a flaw, but what you hear is that you dream big.
It says both things
Look at the extrovert sentence again. At times you are outgoing, at other times reserved. That describes every person who has ever lived. Type descriptions are full of these hinges: you value tradition but also crave new experiences, you can be the life of the party but need time alone afterward. Whatever you are like, half the sentence fits, and your memory supplies an example to go with it.
It is addressed to you
Studies in the same review found that people accept a description more when they are told it came from their own assessment. The same text, presented as true of people in general, persuades less (Dickson and Kelly, 1985). Every test result works this way. Your name sits at the top and your type is printed in large letters.
Authority matters less than you would think
You might expect a description from a psychologist to be more convincing than one from a stranger. Mostly it isn't. The review found that the status of the person giving the feedback had little effect overall, with one exception: unflattering feedback was accepted more readily from someone seen as high status (Dickson and Kelly, 1985). Flattery needs no credentials. Criticism does.
Your chatbot does it too
People now ask ChatGPT to describe their personality from their chat history and pass the answers around as screenshots. Those answers have the same problem as Forer's sketch, and one more on top.
AI assistants are trained partly on which answers people prefer, and people like being agreed with. A 2024 study tested five leading AI assistants and found that all of them consistently drifted toward telling users what they wanted to hear, a habit the researchers called sycophancy. Both people and the models used to train the assistants sometimes preferred a convincingly written agreeable answer over a correct one (Sharma et al., 2024).
So when you ask a chatbot who you are, you are asking a system that leans toward pleasing you to describe the subject you are most pleased to hear about.
Can a language model read personality at all? A little. When researchers had GPT models guess people's Big Five traits from their social media posts, the guesses correlated about .29 with the scores people gave themselves on a questionnaire (Peters and Matz, 2024). A correlation of 1 would be perfect agreement and 0 would be none. So there is a real signal, and it is far too weak to trust any single sentence the model writes about you.
The better test can lose the feel-good contest
Here is the uncomfortable part for anyone who sells personality tests. The research group Clearer Thinking had people take a Jungian type test, the four-letter kind, and a Big Five test, then read a report from each. People liked the type reports more. Only 10% disagreed that the Jungian report made them feel good, against 19% for the Big Five report (Clearer Thinking, 2024).
In the same research, Big Five scores predicted people's life outcomes about twice as well as a four-letter type did (Clearer Thinking, 2024). The report that felt better was the one that told people less.
Types also wobble. Pittenger reported that about half of people land on a different type when they retake the test after only five weeks (Pittenger, 1993, as cited in a 2017 review). The letters that felt so right in March can turn into different letters that feel just as right in April.
The feeling of being seen and the accuracy of a result are separate things, and they can point in opposite directions. If you want the longer version of what the research says about reliability, our guide on how accurate personality tests are walks through it, and Big Five vs MBTI compares the two frameworks directly.
Six checks to run on any description of you
You can't switch the Barnum effect off. You can catch it. These work on a type report, a paid report, a horoscope or a chatbot's answer.
- Do the swap test. Imagine the description was handed to you with a friend's name on top. If you would nod and say that sounds like them, it is too general to tell you anything.
- Look for the cost. An accurate description should say something about you that you would rather it didn't, and that people who know you would recognize. If every line is a compliment, you are reading a Forer sketch.
- Cross out the hinges. Strike every sentence that says you are one way at times and the opposite way at others. Read what is left.
- Ask someone who knows you. You are usually the better judge of traits that live inside your head, like how anxious you feel. People close to you are often better judges of traits that show from the outside, like how intelligent you come across (Vazire, 2010). Hand them the description and ask which lines are wrong.
- Make a prediction. Turn one line into something you can check. "You avoid conflict" predicts that you will put off the awkward message this week. Then watch what you do.
- Retake it in a month. Scores that shift a little are normal. A completely different type is a warning sign about the test, and not a sign that you changed.
A prompt that makes a chatbot work harder
If you ask an AI assistant about yourself, paste this in first:
Describe my personality based on what you know about me. For every claim, quote or paraphrase something I actually said that supports it. Leave out any sentence that would be true of most people. Include at least two traits that most people would see as weaknesses, and say how confident you are in each claim.
The answer will be shorter and less pleasant. It will also give you something you can check.
How we try to pass our own test
We sell a personality report, so this post is a test of us too.
Our free test scores you on continuous scales instead of sorting you into a type. It uses short forms of Big Five items from the public International Personality Item Pool. These are our own short forms. They are not independently validated, and this is a self-reflection tool, not a clinical one.
The paid report is written from your scores, the choices you make in short scenarios and what you write in your own words. It names three blind spots, each meant to be specific and a little uncomfortable, and at least one has to draw on something you wrote. Its 30-day plan starts from the growth area that matters most for you, which means it is supposed to include things you would rather not read.
None of that makes us immune. A report with your name on it triggers the same instincts Forer's students had. So run the six checks on ours as well. You can take the free test and read your trait scores before deciding whether a report is worth paying for. If it fails any of the checks, tell us at hello@depthprofile.com.
Sources
- Forer (1949), The fallacy of personal validation, Journal of Abnormal and Social Psychology
- Meehl (1956), Wanted, a good cookbook, American Psychologist
- Dickson and Kelly (1985), The Barnum effect in personality assessment, Psychological Reports
- Sharma et al. (2024), Towards understanding sycophancy in language models, ICLR
- Peters and Matz (2024), Large language models can infer psychological dispositions of social media users, PNAS Nexus
- Clearer Thinking (2024), How accurate are popular personality test frameworks at predicting life outcomes
- Pittenger (1993), Measuring the MBTI and coming up short, as cited in Acta de Investigacion Psicologica (2017)
- Vazire (2010), Who knows what about a person? Journal of Personality and Social Psychology