Guide
Can AI tell your personality?
Partly. Research shows computers can estimate Big Five traits from digital footprints like social media likes and the way people write, sometimes about as well as people who know them. The predictions are rough, though, and they raise real privacy questions.
Personality leaves traces online
The Big Five show up in behavior, and a lot of behavior now happens online. The pages people like, the words they use, and how they post all carry faint signals about their traits. Extraverts, for example, tend to write more about social events and use more words about people, while more open people tend to engage with art and ideas.
A well-known study from 2015 found that a computer model using Facebook likes could predict people's self-reported Big Five scores about as well as their friends could, and better with more data.
Language models and personality
Large language models, the systems behind chatbots, have made this more powerful because they can analyze free-form text. Researchers are studying how well they can infer personality from what people write, with promising but still imperfect results.
Researchers have also turned Big Five tests on the chatbots themselves. A study led by the University of Cambridge and Google DeepMind found that chatbots respond to personality questionnaires in consistent, human-like ways, and that their apparent personality can be shaped with prompts, for example made to seem more extraverted, with effects that carry over into what they write.
Do chatbots have a personality?
Not in the way people do. A chatbot's apparent personality comes from the text it learned from and how it has been instructed to behave, not from temperament or life experience. That's why it can be adjusted with a few lines of instructions. Still, these measurable styles matter, because they shape how people experience talking to AI.
The limits
- Predictions are estimates. They work reasonably well on average across many people but can be well off for any one person.
- Online behavior is a partial picture. People curate what they post, and platforms shape what they see and do.
- Most models are trained on particular groups and languages and may not work as well for others.
- Traits inferred from behavior are not the same as how you see yourself, which is what a questionnaire measures.
Why privacy matters here
If personality can be inferred without asking, it can be used without consent, for example to target ads or political messages at people's likely traits. That's one reason it's worth knowing what your data can reveal, and why a questionnaire you choose to take, whose results you control, is a very different thing from being profiled.
For understanding yourself, answering the questions honestly is still the most direct and transparent way to get your Big Five profile.
The five traits
More guides
Common questions
Can ChatGPT take a personality test?
It can answer the questions, and researchers have done exactly that. The results reflect the model's training and instructions rather than a personality in the human sense, and they can be changed with prompts.
Is AI better than a personality test?
Not for knowing yourself. Algorithms estimate traits from indirect signals and can be wrong for individuals. A validated questionnaire asks you directly and lets you see exactly how your scores were produced.
Can apps guess my personality without asking?
Research shows it's possible to estimate traits from things like likes and writing, at least roughly. How much any particular app does this depends on its data and practices, so it's worth checking privacy settings and policies.
For self-reflection and education only. This isn't a clinical or diagnostic tool and isn't a substitute for professional psychological or medical advice.